Datasets:
Upload folder using huggingface_hub
Browse files- README.md +89 -0
- codebook.md +62 -0
- data/threads.csv +0 -0
- scripts/agreement.py +50 -0
- scripts/build_dataset.py +93 -0
- scripts/code.py +566 -0
- scripts/collect.py +161 -0
- scripts/filters.py +44 -0
- scripts/refresh.sh +24 -0
- scripts/repos_full.json +995 -0
- scripts/sample_for_review.py +22 -0
- scripts/stats.py +74 -0
README.md
CHANGED
|
@@ -1,3 +1,92 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pretty_name: "AI App Cloud Deploy Issues (GitHub)"
|
| 6 |
+
size_categories:
|
| 7 |
+
- 1K<n<10K
|
| 8 |
+
task_categories:
|
| 9 |
+
- text-classification
|
| 10 |
+
tags:
|
| 11 |
+
- deployment
|
| 12 |
+
- llm-apps
|
| 13 |
+
- devops
|
| 14 |
+
- github-issues
|
| 15 |
+
- ai-generated-code
|
| 16 |
+
configs:
|
| 17 |
+
- config_name: default
|
| 18 |
+
data_files:
|
| 19 |
+
- split: train
|
| 20 |
+
path: data/threads.csv
|
| 21 |
---
|
| 22 |
+
|
| 23 |
+
# AI App Cloud Deploy Issues (GitHub)
|
| 24 |
+
|
| 25 |
+
**What this is:** a coded index of public GitHub issue threads where people try to get an **AI app** (a chatbot, a RAG app, an agent, a model server, or an app generated by an AI builder) running on a **remote host** (Vercel, Azure, AWS, Kubernetes, Hugging Face Spaces, Railway…).
|
| 26 |
+
Each row is one thread: a link, its metadata (dates, counts, state), and columns we coded from the text: what kind of thread it is (problem, feature request, docs request, question, CI), host, app type, framework, model provider, topic (what the thread is about: a failure, a feature or a question), error type, how it ended.
|
| 27 |
+
|
| 28 |
+
**Who made it:** Hal9. We build Cloudish, a hosting service for apps that AI agents deploy, so we have an interest in this topic. The dataset is meant to be useful on its own: it does not mention Cloudish, and Cloudish is not one of the hosts in it.
|
| 29 |
+
|
| 30 |
+
**No text, no authors.** The files contain **no issue text, no titles, no excerpts and no usernames** (not even hashed). The text belongs to its authors and stays on GitHub: follow the `url` to read a thread.
|
| 31 |
+
|
| 32 |
+
## Quick start
|
| 33 |
+
|
| 34 |
+
```python
|
| 35 |
+
from datasets import load_dataset
|
| 36 |
+
|
| 37 |
+
ds = load_dataset("hal9/ai-app-cloud-deploy-issues", split="train") # repo id is a placeholder
|
| 38 |
+
df = ds.to_pandas()
|
| 39 |
+
print(df.groupby("target_host")["resolved"].apply(lambda s: (s == "yes").mean()).sort_values())
|
| 40 |
+
```
|
| 41 |
+
|
| 42 |
+
## Columns
|
| 43 |
+
|
| 44 |
+
See [`codebook.md`](codebook.md) for every column, its values and the rule behind it. In short:
|
| 45 |
+
|
| 46 |
+
- **metadata** (from the GitHub API): `url`, `repo`, `repo_group`, `created_at`, `closed_at`, `state`, `state_reason`, `n_comments`, `n_reactions`, `time_to_first_response_h`, `time_to_close_h`, `maintainer_replied`
|
| 47 |
+
- **coded**: `thread_type`, `target_host`, `app_type`, `framework`, `stack`, `model_provider`, `topic`, `error_type`, `resolution_type`, `resolved`, `ai_generated_app`, `ai_tool`
|
| 48 |
+
- **provenance**: `coding_method` (`keyword` or `keyword+agent_reviewed`), `coding_confidence` (`title`, `body`, `none`: where the error or topic keywords were found), `relevance_method` (`keyword` or `agent_reviewed`), `collected_on`
|
| 49 |
+
|
| 50 |
+
## How it was built
|
| 51 |
+
|
| 52 |
+
1. **Sources.** Public GitHub issues only, created since 2024-01-01, from 34 repositories chosen for precision: AI app frameworks and apps (Gradio, Chainlit, LangGraph, Vercel AI SDK, Flowise, Langflow, Open WebUI, Dify, TGI, vLLM, Ollama…), cloud vendors' AI sample apps (Azure, Google Cloud), and AI app builders / coding agents (bolt.diy, Dyad, Claude Code). The list is in `scripts/repos_full.json`.
|
| 53 |
+
In a pilot, issue trackers of deploy CLIs (Vercel, Netlify, Fly, Railway) gave **0%** relevant threads about AI apps, and LangChain (Python) and LlamaIndex about 21%: they were dropped. SkyPilot (GPU jobs rather than app deploys) and replicate/cog (2 threads) were dropped after the full-build review.
|
| 54 |
+
2. **Search.** One GitHub search query per (repository, title term), for example `repo:Chainlit/chainlit is:issue "cloud run" in:title`. The terms are deploy words (`deploy`, `production`, `hosting`) and host names.
|
| 55 |
+
3. **Caps.** At most 2 threads per author across the whole dataset (the author is used for this cap during collection, then dropped), and at most 250 threads per repository.
|
| 56 |
+
4. **Relevance filter.** Nine written rules decide what "deploying or running an AI app on a host" means (see the codebook): a host is a remote machine (not a laptop, WSL or a local GPU); model serving on a host counts; an AI agent that deploys and hits its own limits counts; an app bug that merely shows up on a server does not; calling a cloud's AI API without deploying an app does not; in general frameworks (Streamlit, Gradio) the thread itself must show an AI app; CI of a cloud integration counts. Keyword rules apply them; every row the keywords could not settle (about 220) was read and judged by the AI assistant. Feature and docs requests are kept and labelled by `thread_type`.
|
| 57 |
+
5. **Coding.** Keyword rules (`scripts/code.py`), with no paid API. A sample of threads was then read and coded by the AI assistant (Claude) that built the dataset; its codes replace the keyword codes for those rows (`coding_method = keyword+agent_reviewed`). The relevance rules were set with a person from our team, who also checked a small sample of rows against the threads; the quality figures below come from the larger samples the AI reviewer read.
|
| 58 |
+
|
| 59 |
+
## Quality (measured, not assumed)
|
| 60 |
+
|
| 61 |
+
On samples read thread by thread by the reviewer (see step 5):
|
| 62 |
+
|
| 63 |
+
- **relevance precision** of rows decided by keywords alone (`relevance_method = keyword`): **70%** (42 of a fresh random sample of 60, drawn after the last rule change; 95% Wilson interval 58–80%). Counting the rows the reviewer read (`agent_reviewed`, about 28% of the file) as correct, the whole file is about **79%** relevant (about 70–86%). Most false positives are an app bug or an API call that mentions a deployment in passing, or a local install. These figures come from the AI reviewer, who also wrote the keyword rules
|
| 64 |
+
- **thread type** (`thread_type`) agrees with the reviewer on 49 of the same 60 rows (82%); the keyword `other` is the weak value
|
| 65 |
+
- **agreement of keyword codes with reviewed codes**: host 0.95–0.98, app type 0.95, framework 0.95–0.98, model provider 0.83–0.85, topic 0.48–0.70, error type 0.57–0.80, resolution 0.65–0.85 (two samples of 46 and 20 threads; the higher figure comes from a sample where the reviewer could see the keyword codes, so it is probably optimistic). The topic and error type figures were measured before three topics (`agent_behavior`, `observability`, `performance_scaling`) and four `agent_*` error types were added; those only relabel rows that were `other`, and their precision was checked separately: 28/28, 17/18 and 17/19 rows read.
|
| 66 |
+
|
| 67 |
+
`topic` and `error_type` are the weakest columns. Filter on `coding_confidence == "title"` when you need them to be reliable.
|
| 68 |
+
|
| 69 |
+
## Limits and biases
|
| 70 |
+
|
| 71 |
+
- **Population.** These are GitHub issue filers: mostly developers, reporting to a framework or sample maintainer. People who build with Lovable, Bolt.new or v0 and ask for help on Discord or Reddit are not here (those sources forbid automated collection or AI use).
|
| 72 |
+
- **Repository mix drives the numbers.** Azure has many threads because two Azure sample apps have busy trackers, not because Azure is harder. Compare hosts within a repo group, not across the whole file.
|
| 73 |
+
- **Search ranking.** GitHub's best-match ranking plus the per-repo cap decides what is in. It is not a random sample of all issues.
|
| 74 |
+
- **Resolution is a best guess.** `resolved` uses GitHub's `state_reason` and the author's own "it works now" messages. Many trackers auto-close stale issues; we try to detect it, but not always.
|
| 75 |
+
- **Language.** Mostly English; some Chinese (NextChat, Dify) and a few Portuguese and Spanish threads.
|
| 76 |
+
- **Thread types.** 66% of rows are problems, 13% feature requests, 12% questions, 4% docs requests, 4% other, under 1% CI. Filter on `thread_type == "problem"` for failure statistics.
|
| 77 |
+
- **Snapshot.** Counts and states are as of `collected_on`.
|
| 78 |
+
|
| 79 |
+
## Legal and ethics
|
| 80 |
+
|
| 81 |
+
- Collected through the GitHub REST API, within rate limits, read-only.
|
| 82 |
+
- GitHub's Acceptable Use Policy (section 7) lets researchers use public, non-personal information for research, provided resulting publications are open access. This dataset and its article are published openly. We store no personal information: no usernames, no names, no emails, no text.
|
| 83 |
+
- The issue text stays under its authors' rights on GitHub; we only link to it.
|
| 84 |
+
- **Removal:** if you filed one of these issues and want its row removed, open a discussion on this dataset page and we will remove it.
|
| 85 |
+
|
| 86 |
+
## License
|
| 87 |
+
|
| 88 |
+
Our coded data (the columns we produced) and the scripts: **CC BY 4.0**. The linked GitHub content is not included and is not covered by this license.
|
| 89 |
+
|
| 90 |
+
## Citation
|
| 91 |
+
|
| 92 |
+
Hal9 (2026). *AI App Cloud Deploy Issues (GitHub)*. Hugging Face dataset. The scripts, the coding rules and the sample review were made with AI help (Claude) in October 2026.
|
codebook.md
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Codebook: AI app cloud-deploy threads (GitHub issues)
|
| 2 |
+
|
| 3 |
+
One row = one public GitHub issue thread (issue + its comments) about deploying an AI app to a remote host.
|
| 4 |
+
Published rows contain **no text and no author information**: only the link, metadata and our coded columns.
|
| 5 |
+
Coded columns come from keyword rules (`scripts/code.py`). Some rows were read thread by thread by the AI assistant (Claude) that built the dataset: the review samples, and every row the keywords could not settle under the relevance rules (v4). Its judgements replace the keyword values (`relevance_method = agent_reviewed`, `coding_method = keyword+agent_reviewed`). No person has checked these rows yet.
|
| 6 |
+
|
| 7 |
+
## Identification and metadata (from the GitHub API)
|
| 8 |
+
|
| 9 |
+
| column | type | definition |
|
| 10 |
+
|---|---|---|
|
| 11 |
+
| `url` | string | link to the issue on github.com (the only way to read the text, which belongs to its authors) |
|
| 12 |
+
| `source` | string | always `github_issues` in this version |
|
| 13 |
+
| `repo` | string | `owner/name` of the repository |
|
| 14 |
+
| `repo_group` | enum | why the repo was searched: `host` (deploy platform CLI), `framework` (AI app framework or AI app), `framework_generic` (general framework: Streamlit, Gradio; the thread itself must show an AI app), `sample` (cloud vendor AI sample app), `builder` (AI app builder or coding agent) |
|
| 15 |
+
| `created_at` | ISO 8601 UTC | issue creation |
|
| 16 |
+
| `closed_at` | ISO 8601 UTC or empty | last close time |
|
| 17 |
+
| `state` | enum | `open` / `closed` at collection time |
|
| 18 |
+
| `state_reason` | enum or empty | GitHub's own value: `completed`, `not_planned`, `duplicate`, `reopened`, empty |
|
| 19 |
+
| `n_comments` | int | comment count at collection time |
|
| 20 |
+
| `n_reactions` | int | total reactions on the issue body |
|
| 21 |
+
| `time_to_first_response_h` | float or empty | hours from creation to the first comment by someone other than the author, bots excluded; empty = no such comment |
|
| 22 |
+
| `time_to_close_h` | float or empty | hours from creation to `closed_at` |
|
| 23 |
+
| `maintainer_replied` | yes/no | at least one comment by a user whose `author_association` is OWNER, MEMBER or COLLABORATOR (other than the author) |
|
| 24 |
+
| `collected_on` | date | snapshot date |
|
| 25 |
+
|
| 26 |
+
## Coded columns
|
| 27 |
+
|
| 28 |
+
| column | values | rule (summary) |
|
| 29 |
+
|---|---|---|
|
| 30 |
+
| `thread_type` | `problem`, `feature`, `docs`, `question`, `pr_ci`, `other` | what the thread is. `problem`: something breaks or blocks; `feature`: asks for a new capability; `docs`: asks for or fixes documentation; `question`: how-to, nothing broken yet; `pr_ci`: a code change, release or CI/test pipeline (issues only in this version, so "PR" means an issue about one); `other`: none of these (measurements, discussions). Rule, first match wins: a title tag (`[CI]`, `test:`, `[Docs]`, `docs:`, `[Feature]`, `feat:`, `[RFC]`, `[Question]`, `[Usage]`, `[Bug]`…); CI/test words in the title or a CI label; docs words in the title (docs, documentation, README, guide, tutorial) with no error word, or a documentation label without a bug label; a feature label (`enhancement`, `feature`, `rfc`, `integration request`) or a feature title pattern (`Add support…`, `Support for…`, `Allow…`, `option to…`) with no error word; an error word (error, fails, broken, cannot, 4xx/5xx, timeout…) or a bug label; a question pattern (`How to…`, `Is it possible…`, `…?`) or a question/support label; an error word in the body; else `other`. Agreement with the reviewer on a fresh sample: 49/60 (82%); most misses are keyword `other` that the reviewer reads as `feature` or `question` |
|
| 31 |
+
| `target_host` | `vercel`, `netlify`, `cloudflare`, `fly`, `railway`, `render`, `heroku`, `supabase`, `hf_spaces`, `hf_inference_endpoints`, `streamlit_cloud`, `posit_connect`, `langgraph_platform`, `azure`, `gcp`, `aws`, `digitalocean`, `runpod`, `modal`, `replicate`, `kubernetes` (generic, any cloud), `vps_docker` (self-managed server, Docker, Coolify-like panels), `unspecified` | first host named in the title, else in the body, else the repo's own platform (e.g. `vercel/vercel` → `vercel`, Azure samples → `azure`) |
|
| 32 |
+
| `app_type` | `chatbot`, `rag`, `ai_generated_web_app`, `model_serving`, `agent`, `other` | agent cues > RAG cues > builder repo > serving cues > chat cues; repo defaults for single-purpose repos |
|
| 33 |
+
| `framework` | one of `streamlit`, `gradio`, `chainlit`, `nextjs`, `remix`, `sveltekit`, `nuxt`, `vite_react`, `astro`, `fastapi`, `flask`, `django`, `express_node`, `hono`, `langgraph`, `langchain`, `llamaindex`, `vercel_ai_sdk`, `flowise`, `langflow`, `open_webui`, `nextchat`, `chatbot_ui`, `cog`, `tgi`, `ollama`, `vllm`, `dify`, `ragflow`, `anythingllm`, `hf_chat_ui`, `litellm`, `mastra`, `google_adk`, `crewai`, `agno`, `skypilot`, `prompty`, `unspecified` | the repo's framework if it is one, else first framework named. In builder repos the framework of the generated app is usually not stated, so `unspecified`; self-hosted bolt.diy is `remix` |
|
| 34 |
+
| `stack` | `;`-list of the values above | every framework named (repo framework first) |
|
| 35 |
+
| `model_provider` | `;`-list (max 3) of `azure_openai`, `openai`, `anthropic`, `google`, `aws_bedrock`, `local_ollama`, `huggingface`, `mistral`, `deepseek`, `groq`, `openrouter`, `replicate`, `cohere`, or `unspecified` | providers named in title or body; repo default for single-provider samples |
|
| 36 |
+
| `topic` | `build`, `runtime_dependencies`, `provisioning_iac` (cloud template / infrastructure provisioning, e.g. `azd up`, Bicep, Terraform), `memory_model_size`, `secrets_api_keys`, `gpu`, `costs_quota`, `cold_start`, `timeouts`, `database`, `auth_permissions`, `networking`, `streaming`, `routing_config`, `agent_behavior` (a coding or deploy agent itself causes the problem: deploys unasked, claims a deploy worked, deploys to the wrong target, or is blocked by its own safety check; `error_type` says which), `observability` (monitoring, tracing and telemetry tooling: Langfuse, OpenTelemetry/OTLP, Sentry, Prometheus, Grafana, Datadog, metrics, ServiceMonitor; plain logs do not count), `performance_scaling` (slowness, latency, throughput, concurrency, autoscaling, replicas, high availability), `other` | what the thread is about, whatever its `thread_type` (a failure, a feature request, a question or a docs request). Called `problem_category` before 2026-10-30. From the `error_type` found in the title; else a topical cue in the title (GPU, cold start, database, auth, cost); else the `error_type` found in the body; `coding_confidence` says which. Only when all of these give `other`: the three topics added on 2026-10-30, title first, then body (`agent_behavior` needs an agent context: the Claude Code repo, or the text names an agent such as Claude, Codex, Copilot, Cursor; `observability` and `performance_scaling` use stricter terms in the body, on its text without code blocks). So they never take a row from the 14 older topics, and `coding_confidence` does not change for them (it stays `none` when no error or older cue was found). Else `other` |
|
| 37 |
+
| `error_type` | `secret_exposed` (credentials deployed or published by mistake: a credential word such as secret, API key, password, `.env` must sit next to a leak/expose/publish word; memory leaks are `oom`, exposed ports `port_binding`), `iac_template_error`, `oom`, `size_limit` (incl. disk/volume full), `timeout`, `missing_env_var`, `quota_billing`, `permission_rbac`, `port_binding`, `module_dependency`, `build_failure`, `cors_network`, `streaming`, `not_found_routing`, `server_error_5xx`, `agent_deployed_unasked` (the agent deploys or pushes to production without permission or confirmation, or against instructions), `agent_false_success` (the agent says a deploy worked, was verified or happened when it did not), `agent_wrong_target` (the agent deploys to the wrong project, environment, account, repo or branch), `agent_blocked_refused` (an auto-mode classifier, safety filter or the model blocks or refuses a legitimate, user-authorised deploy: a false positive), `other_unknown` | first matching pattern in the title, else in the first 3,000 characters of the body. For provisioning failures the inner cause (quota, permission) wins when stated. The four `agent_*` types (added 2026-10-30) come with `topic = agent_behavior` only; they replace `other_unknown`, or a `server_error_5xx` found only in the body |
|
| 38 |
+
| `resolution_type` | `unanswered`, `self_resolved`, `config_fix`, `version_upgrade`, `version_downgrade`, `workaround`, `platform_limit_no_fix`, `user_error_config`, `duplicate`, `stale_closed`, `open_unresolved`, `closed_unclear` | from replies (non-author, non-bot), labels and `state_reason` |
|
| 39 |
+
| `resolved` | yes/no | closed as `completed` (not by a stale bot), or the author says it works / is solved |
|
| 40 |
+
| `ai_generated_app` | yes/no/unknown | `yes` if the thread is in an AI app builder or coding-agent repo, or the text says the app was built/generated with an AI tool. `no` only when the thread is about self-hosting the builder itself (e.g. deploying bolt.diy). Otherwise `unknown`: people rarely say they wrote the app by hand |
|
| 41 |
+
| `ai_tool` | `lovable`, `bolt`, `v0`, `cursor`, `claude_code`, `replit`, `copilot`, `windsurf`, `dyad`, `codex`, or empty | tool named, or repo default |
|
| 42 |
+
| `coding_method` | `keyword` / `keyword+agent_reviewed` | `keyword+agent_reviewed`: the AI reviewer read the thread and corrected the codes (host, app type, framework…) |
|
| 43 |
+
| `coding_confidence` | `title` / `body` / `none` | where the error keywords or the older topic cues were found; `title` rows agree best with the review. The three topics added 2026-10-30 do not set it |
|
| 44 |
+
| `relevance_method` | `keyword` / `agent_reviewed` | who decided the row is relevant (and its `thread_type`): the keyword rules, or the AI reviewer (Claude), not a person |
|
| 45 |
+
|
| 46 |
+
## Relevance (rules v4, decided by the dataset owner on 2026-10-28)
|
| 47 |
+
|
| 48 |
+
A thread is **relevant** if it is about deploying or running an AI app on a host (cloud AND AI). Relevance is `yes` or `no`; the earlier `adjacent` level is gone, and feature and docs requests stay in, told apart by `thread_type`.
|
| 49 |
+
|
| 50 |
+
1. **Repo context counts.** A Kubernetes or Docker thread in an AI app's repo (Dify, Langflow…) is relevant even if the thread never mentions AI.
|
| 51 |
+
2. **A host is a remote machine:** public cloud, VPS, company server, Kubernetes. Docker on a laptop, WSL or a local GPU is not.
|
| 52 |
+
3. **Model serving on a host** (vLLM, TGI, Ollama on a remote GPU, a model deployed on Azure) is relevant; `app_type = model_serving`.
|
| 53 |
+
4. **An agent that deploys and hits its own limits** (Claude Code blocked by a classifier, permission denied, a false "it's deployed") is relevant; `ai_generated_app = yes`.
|
| 54 |
+
5. **Feature and docs requests stay in**, with `thread_type` saying what they are.
|
| 55 |
+
6. **An app bug seen on a server but not caused by the deploy, the config or the host runtime is not relevant** (a model looping on a k8s GPU, a UI bug after an update). There must be a link to the host.
|
| 56 |
+
7. **Calling a cloud's managed AI API without deploying your own app is not relevant** (Vertex batch, Bedrock, Azure OpenAI calls, "deployment name" settings, LiteLLM router "deployments").
|
| 57 |
+
8. **General frameworks** (Streamlit, Gradio, Next.js, host CLIs): the thread must show an AI app (model, LLM, RAG…); the repo alone is not enough. Rule 1 applies to AI-specific repos only.
|
| 58 |
+
9. **Tests or CI of a framework's cloud integration** are relevant; `thread_type = pr_ci`.
|
| 59 |
+
|
| 60 |
+
**How the rules are applied.** Keywords first (`relevance()` in `scripts/code.py`): the title must name a hosting platform or a deploy verb, plus the exclusions above (local cues, managed-API names without a hosting term, router senses of "deployment", CI without a cloud named, Claude Code threads about the desktop app, browser pane, OAuth or connectors). Where the keywords cannot settle a row they set `kw_unsure` (private file only): Claude Code threads whose deploy context is unclear, model-serving threads that name no host, app-behaviour titles with no host cue (rule 6), threads whose only host cue is "in production", managed-API titles that also say "deploy". Every such row was judged by the AI reviewer (`manual/judgement_v4.csv`).
|
| 61 |
+
|
| 62 |
+
Only rows coded relevant are published.
|
data/threads.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
scripts/agreement.py
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Agreement between keyword coding and manual judgement.
|
| 3 |
+
Usage: python3 -I scripts/agreement.py coded_kw.jsonl manual/judgement.csv
|
| 4 |
+
judgement.csv: idx (row order in coded file) or url, relevant (yes/adjacent/no), then optional
|
| 5 |
+
manual codes; empty code = not coded by hand (relevance only)."""
|
| 6 |
+
import csv, json, sys, collections
|
| 7 |
+
|
| 8 |
+
coded = [json.loads(l) for l in open(sys.argv[1])]
|
| 9 |
+
by_url = {c["url"]: c for c in coded}
|
| 10 |
+
rows = list(csv.DictReader(open(sys.argv[2])))
|
| 11 |
+
FIELDS = ["target_host", "app_type", "framework", "model_provider", "topic",
|
| 12 |
+
"error_type", "resolution_type", "ai_generated_app"]
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
def get(r):
|
| 16 |
+
if r.get("url"):
|
| 17 |
+
return by_url.get(r["url"])
|
| 18 |
+
return coded[int(r["idx"])]
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
rel = collections.Counter()
|
| 22 |
+
for r in rows:
|
| 23 |
+
c = get(r)
|
| 24 |
+
if c is None:
|
| 25 |
+
continue
|
| 26 |
+
rel[(c["kw_relevant"], r["relevant"])] += 1
|
| 27 |
+
n = sum(rel.values())
|
| 28 |
+
print(f"relevance (kw, manual) on {n}: {dict(rel)}")
|
| 29 |
+
yes_kw = sum(v for (k, m), v in rel.items() if k == "yes")
|
| 30 |
+
print(f" precision of kw 'yes' vs manual yes: {sum(v for (k, m), v in rel.items() if k == 'yes' and m == 'yes')}/{yes_kw}")
|
| 31 |
+
man_yes = sum(v for (k, m), v in rel.items() if m == "yes")
|
| 32 |
+
print(f" recall of kw 'yes' over manual yes: {rel[('yes', 'yes')]}/{man_yes}")
|
| 33 |
+
print(f" manual precision of the collected set: {man_yes}/{n} = {man_yes / n:.2f}")
|
| 34 |
+
|
| 35 |
+
coded_rows = [r for r in rows if r["relevant"] == "yes" and r.get("app_type")]
|
| 36 |
+
print(f"\nfield agreement on {len(coded_rows)} hand-coded relevant threads:")
|
| 37 |
+
for f in FIELDS:
|
| 38 |
+
ok = tot = 0
|
| 39 |
+
for r in coded_rows:
|
| 40 |
+
c = get(r)
|
| 41 |
+
kw = c["kw_" + f] if "kw_" + f in c else c.get(f)
|
| 42 |
+
if f == "topic" and kw is None: # coded files from before 2026-10-30
|
| 43 |
+
kw = c.get("kw_problem_category")
|
| 44 |
+
man = r.get(f) if f in r else r.get("problem_category", "") # judgement files keep the old column name
|
| 45 |
+
if f == "model_provider": # compare as sets
|
| 46 |
+
ok += set((kw or "").split(";")) == set(man.split(";"))
|
| 47 |
+
else:
|
| 48 |
+
ok += (kw or "") == man
|
| 49 |
+
tot += 1
|
| 50 |
+
print(f" {f:18s} {ok}/{tot} = {ok / tot:.2f}")
|
scripts/build_dataset.py
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Merge keyword coding + manual judgements into the publishable CSV and the private verification file.
|
| 3 |
+
|
| 4 |
+
Usage:
|
| 5 |
+
python3 -I scripts/build_dataset.py CODED.jsonl OUT_PUBLIC.csv OUT_PRIVATE.csv [--all] [JUDGEMENT.csv ...]
|
| 6 |
+
JUDGEMENT.csv rows are keyed by `url`, or by `idx` = row number in raw/coded_kw.jsonl (pilot file).
|
| 7 |
+
Files are merged field by field in the order given: a non-empty value in a later file wins.
|
| 8 |
+
Pass manual/judgement_v4.csv last: it holds the relevance (yes/no) and thread_type under rules v4.
|
| 9 |
+
The judgement files call the topic column `problem_category` (its name before 2026-10-30); it is read as `topic`.
|
| 10 |
+
A judged topic `other` predates the topics added 2026-10-30 (agent_behavior, observability,
|
| 11 |
+
performance_scaling): when the keywords give one of those, the keyword value is kept (same for an agent_*
|
| 12 |
+
error_type over a judged other_unknown).
|
| 13 |
+
All judgements so far are by the AI assistant (Claude), not by a person (column judged_by).
|
| 14 |
+
--all : write every thread with a `relevance` column (pilot.csv); default: only relevant == yes.
|
| 15 |
+
Published file: NO text, NO author info (no title, no excerpt, no author hash).
|
| 16 |
+
Private file : url, title, 30-word excerpt, codes — local verification only, never published.
|
| 17 |
+
"""
|
| 18 |
+
import csv, json, os, sys
|
| 19 |
+
from datetime import date
|
| 20 |
+
|
| 21 |
+
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
| 22 |
+
args = [a for a in sys.argv[1:] if a != "--all"]
|
| 23 |
+
ALL = "--all" in sys.argv
|
| 24 |
+
coded_path, out_pub, out_priv, judg_paths = args[0], args[1], args[2], args[3:]
|
| 25 |
+
coded = [json.loads(l) for l in open(coded_path)]
|
| 26 |
+
pilot = [json.loads(l) for l in open(os.path.join(ROOT, "raw", "coded_kw.jsonl"))]
|
| 27 |
+
|
| 28 |
+
manual = {}
|
| 29 |
+
for jp in judg_paths:
|
| 30 |
+
for r in csv.DictReader(open(jp)):
|
| 31 |
+
url = r.get("url") or pilot[int(r["idx"])]["url"]
|
| 32 |
+
m = manual.setdefault(url, {})
|
| 33 |
+
for k, v in r.items():
|
| 34 |
+
if k == "problem_category": # old name of `topic`
|
| 35 |
+
k = "topic"
|
| 36 |
+
if v not in ("", None):
|
| 37 |
+
m[k] = v
|
| 38 |
+
coded_urls = {c["url"] for c in coded}
|
| 39 |
+
for url, m in manual.items():
|
| 40 |
+
if url in coded_urls and m.get("relevant") not in ("yes", "no"):
|
| 41 |
+
sys.exit(f"{url}: relevance must be yes/no under rules v4 (got {m.get('relevant')!r}); re-judge it in manual/judgement_v4.csv")
|
| 42 |
+
|
| 43 |
+
CODES = ["target_host", "app_type", "framework", "model_provider", "topic", "error_type",
|
| 44 |
+
"resolution_type", "ai_generated_app", "ai_tool"]
|
| 45 |
+
HAND = ["target_host", "app_type", "framework", "model_provider", "topic", "error_type", "resolution_type"]
|
| 46 |
+
PUB = ["url", "source", "repo", "repo_group", "created_at", "closed_at", "state", "state_reason", "n_comments",
|
| 47 |
+
"n_reactions", "maintainer_replied", "time_to_first_response_h", "time_to_close_h", "resolved", "thread_type"] + \
|
| 48 |
+
CODES[:3] + ["stack"] + CODES[3:] + ["coding_method", "coding_confidence", "relevance_method", "collected_on"]
|
| 49 |
+
if ALL:
|
| 50 |
+
PUB.insert(1, "relevance")
|
| 51 |
+
PRIV = ["url", "title", "excerpt_private", "relevance", "relevance_method", "kw_relevant", "kw_unsure", "manual_note",
|
| 52 |
+
"thread_type", "kw_thread_type"] + CODES + ["coding_method"]
|
| 53 |
+
|
| 54 |
+
NEW_TOPICS = {"agent_behavior", "observability", "performance_scaling"} # added 2026-10-30
|
| 55 |
+
today = date.today().isoformat()
|
| 56 |
+
pub_rows, priv_rows = [], []
|
| 57 |
+
for c in coded:
|
| 58 |
+
m = manual.get(c["url"])
|
| 59 |
+
rel = m["relevant"] if m else c["kw_relevant"]
|
| 60 |
+
row = {k: c.get(k) for k in ["url", "source", "repo", "created_at", "closed_at", "state", "state_reason",
|
| 61 |
+
"n_comments", "n_reactions", "maintainer_replied", "time_to_first_response_h",
|
| 62 |
+
"time_to_close_h", "resolved", "coding_confidence"]}
|
| 63 |
+
row["repo_group"] = c["group"]
|
| 64 |
+
row["stack"] = c.get("kw_stack")
|
| 65 |
+
for f in CODES:
|
| 66 |
+
row[f] = c.get("kw_" + f)
|
| 67 |
+
hand = bool(m and any(m.get(f) for f in HAND))
|
| 68 |
+
if m: # manual values override when given (partial overrides allowed, e.g. ai_generated_app only)
|
| 69 |
+
for f in CODES:
|
| 70 |
+
if m.get(f):
|
| 71 |
+
if f == "topic" and m[f] == "other" and row[f] in NEW_TOPICS:
|
| 72 |
+
continue
|
| 73 |
+
if f == "error_type" and m[f] == "other_unknown" and (row[f] or "").startswith("agent_"):
|
| 74 |
+
continue
|
| 75 |
+
row[f] = m[f]
|
| 76 |
+
row["thread_type"] = (m or {}).get("thread_type") or c.get("kw_thread_type")
|
| 77 |
+
row["coding_method"] = "keyword+agent_reviewed" if hand else "keyword"
|
| 78 |
+
row["relevance"] = rel
|
| 79 |
+
row["relevance_method"] = "agent_reviewed" if m else "keyword"
|
| 80 |
+
row["collected_on"] = c.get("collected_on", today)
|
| 81 |
+
if ALL or rel == "yes":
|
| 82 |
+
pub_rows.append({k: row.get(k) for k in PUB})
|
| 83 |
+
priv_rows.append({**{k: row.get(k) for k in PRIV}, "title": c["title"], "excerpt_private": c.get("excerpt_private"),
|
| 84 |
+
"kw_relevant": c["kw_relevant"], "kw_unsure": c.get("kw_unsure"), "kw_thread_type": c.get("kw_thread_type"),
|
| 85 |
+
"manual_note": (m or {}).get("note")})
|
| 86 |
+
|
| 87 |
+
for path, rows, cols in ((out_pub, pub_rows, PUB), (out_priv, priv_rows, PRIV)):
|
| 88 |
+
os.makedirs(os.path.dirname(os.path.abspath(path)), exist_ok=True)
|
| 89 |
+
with open(path, "w", newline="") as f:
|
| 90 |
+
w = csv.DictWriter(f, fieldnames=cols)
|
| 91 |
+
w.writeheader()
|
| 92 |
+
w.writerows(rows)
|
| 93 |
+
print(f"{out_pub}: {len(pub_rows)} rows; {out_priv}: {len(priv_rows)} rows")
|
scripts/code.py
ADDED
|
@@ -0,0 +1,566 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Keyword coding of collected threads (no paid APIs, no LLM).
|
| 3 |
+
|
| 4 |
+
Input : raw/threads.jsonl (from collect.py)
|
| 5 |
+
Output: raw/coded_kw.jsonl (all metadata + kw_* coded columns + private excerpt)
|
| 6 |
+
Usage : python3 -I scripts/code.py [threads.jsonl] [out.jsonl] [--drop-removed]
|
| 7 |
+
--drop-removed skips the repos removed after the full-build review (SkyPilot, replicate/cog)
|
| 8 |
+
Rules are ordered: first match wins for single-valued columns; list columns join
|
| 9 |
+
up to 3 matches with ';'. See codebook.md for definitions.
|
| 10 |
+
"""
|
| 11 |
+
import json, os, re, sys
|
| 12 |
+
|
| 13 |
+
HERE = os.path.dirname(os.path.abspath(__file__))
|
| 14 |
+
sys.path.insert(0, HERE)
|
| 15 |
+
import filters # noqa: E402
|
| 16 |
+
|
| 17 |
+
ROOT = os.path.dirname(HERE)
|
| 18 |
+
I = re.I | re.A # ASCII \b so that "netlify现在" still matches
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
def rx(s):
|
| 22 |
+
return re.compile(s, I)
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
HOSTS = [ # (label, regex) — order = priority when counts tie
|
| 26 |
+
("hf_inference_endpoints", rx(r"inference endpoints?\b|ui\.endpoints\.huggingface|dedicated endpoint")),
|
| 27 |
+
("hf_spaces", rx(r"hugging ?face spaces?|\bhf spaces?|\b(on|to|in) (a |the |my )?(hf |hugging ?face )?spaces?\b|spaces? (sdk|hardware|build|container)|huggingface\.co/spaces|hf\.space")),
|
| 28 |
+
("streamlit_cloud", rx(r"streamlit (community )?cloud|share\.streamlit|streamlit\.app\b")),
|
| 29 |
+
("vercel", rx(r"\bvercel\b(?!/ai)(?! ai sdk)")),
|
| 30 |
+
("netlify", rx(r"\bnetlify\b")),
|
| 31 |
+
("cloudflare", rx(r"\bcloudflare|\bworkers?\.dev|wrangler|\bpages\.dev")),
|
| 32 |
+
("fly", rx(r"\bfly\.io\b|\bflyctl\b|\bfly deploy\b|fly\.toml")),
|
| 33 |
+
("railway", rx(r"\brailway\b")),
|
| 34 |
+
("render", rx(r"\brender\.com\b|\bon render\b|\bto render\b|onrender\.com|render\.yaml")),
|
| 35 |
+
("heroku", rx(r"\bheroku\b")),
|
| 36 |
+
("supabase", rx(r"\bsupabase\b")),
|
| 37 |
+
("langgraph_platform", rx(r"langgraph (platform|cloud)|langsmith deploy|langgraph deploy|langgraph-api|langgraph up")),
|
| 38 |
+
("azure", rx(r"\bazure\b|\bazd\b|app service|container apps|\baks\b")),
|
| 39 |
+
("gcp", rx(r"cloud ?run\b|\bgcp\b|google cloud|app engine|\bgke\b|vertex|cloud functions|firebase")),
|
| 40 |
+
("aws", rx(r"\baws\b|aws lambda|\b(on|to|in|via) lambda\b|lambda (deploy\w*|layers?|cold start|timeouts?|function url|@edge)|sagemaker|\becs\b|\beks\b|\bec2\b|amplify|bedrock|fargate|elastic beanstalk|apprunner|app runner")),
|
| 41 |
+
("digitalocean", rx(r"digital ?ocean|\bdroplet")),
|
| 42 |
+
("runpod", rx(r"\brunpod\b")),
|
| 43 |
+
("modal", rx(r"\bmodal\.com\b|\bmodal (deploy|serve|labs)\b|on modal\b")),
|
| 44 |
+
("replicate", rx(r"\breplicate\.com\b|\bcog push\b|on replicate\b|to replicate\b")),
|
| 45 |
+
("kubernetes", rx(r"kubernetes|\bk8s\b|\bhelm\b|openshift")),
|
| 46 |
+
("posit_connect", rx(r"posit connect|rstudio connect")),
|
| 47 |
+
("vps_docker", rx(r"\bvps\b|docker[- ]compose|\bdocker\b|self[- ]host\w*|nginx|own server|ubuntu server|coolify|easypanel|dokploy|caprover")),
|
| 48 |
+
]
|
| 49 |
+
REPO_HOST = {"vercel/vercel": "vercel", "netlify/cli": "netlify", "superfly/flyctl": "fly",
|
| 50 |
+
"railwayapp/cli": "railway", "Azure-Samples/azure-search-openai-demo": "azure",
|
| 51 |
+
"Azure-Samples/chat-with-your-data-solution-accelerator": "azure",
|
| 52 |
+
"microsoft/sample-app-aoai-chatGPT": "azure", "GoogleCloudPlatform/generative-ai": "gcp", "GoogleCloudPlatform/agent-starter-pack": "gcp",
|
| 53 |
+
"Azure-Samples/contoso-chat": "azure", "Azure-Samples/rag-postgres-openai-python": "azure",
|
| 54 |
+
"replicate/cog": "replicate"}
|
| 55 |
+
|
| 56 |
+
FRAMEWORKS = [
|
| 57 |
+
("streamlit", rx(r"\bstreamlit\b")), ("gradio", rx(r"\bgradio\b")), ("chainlit", rx(r"\bchainlit\b")),
|
| 58 |
+
("nextjs", rx(r"next\.?js|\bnext (13|14|15|16)\b|app router|\bnext build\b")),
|
| 59 |
+
("remix", rx(r"\bremix\b")), ("sveltekit", rx(r"svelte")), ("nuxt", rx(r"\bnuxt")),
|
| 60 |
+
("vite_react", rx(r"\bvite\b")), ("astro", rx(r"\bastro\b")),
|
| 61 |
+
("fastapi", rx(r"fastapi|uvicorn")), ("flask", rx(r"\bflask\b")), ("django", rx(r"\bdjango\b")),
|
| 62 |
+
("express_node", rx(r"\bexpress\b|node\.?js")), ("hono", rx(r"\bhono\b")),
|
| 63 |
+
("langgraph", rx(r"langgraph")), ("langchain", rx(r"langchain")), ("llamaindex", rx(r"llama_?index|llama-index")),
|
| 64 |
+
("vercel_ai_sdk", rx(r"ai[- ]sdk|vercel/ai|@ai-sdk|useChat|streamText")),
|
| 65 |
+
("flowise", rx(r"flowise")), ("langflow", rx(r"langflow")), ("open_webui", rx(r"open[- ]?webui")),
|
| 66 |
+
("nextchat", rx(r"nextchat|chatgpt-next-web")), ("chatbot_ui", rx(r"chatbot[- ]ui")),
|
| 67 |
+
("cog", rx(r"\bcog (push|predict|build)\b|cog\.yaml")), ("tgi", rx(r"text-generation-inference|\btgi\b")), ("ollama", rx(r"\bollama\b")),
|
| 68 |
+
("vllm", rx(r"\bvllm\b")), ("dify", rx(r"\bdify\b")), ("ragflow", rx(r"ragflow")), ("anythingllm", rx(r"anything-?llm")),
|
| 69 |
+
("litellm", rx(r"litellm")), ("mastra", rx(r"\bmastra\b")), ("google_adk", rx(r"\badk\b|agent development kit")),
|
| 70 |
+
("crewai", rx(r"crew ?ai")), ("agno", rx(r"\bagno\b")), ("skypilot", rx(r"skypilot|\bsky (serve|launch)")),
|
| 71 |
+
]
|
| 72 |
+
REPO_FRAMEWORK = {"streamlit/streamlit": "streamlit", "gradio-app/gradio": "gradio", "Chainlit/chainlit": "chainlit",
|
| 73 |
+
"langchain-ai/langchain": "langchain", "langchain-ai/langchainjs": "langchain",
|
| 74 |
+
"langchain-ai/langgraph": "langgraph", "vercel/ai": "vercel_ai_sdk", "ollama/ollama": "ollama",
|
| 75 |
+
"ChatGPTNextWeb/NextChat": "nextchat", "mckaywrigley/chatbot-ui": "chatbot_ui",
|
| 76 |
+
"FlowiseAI/Flowise": "flowise", "langflow-ai/langflow": "langflow",
|
| 77 |
+
"run-llama/llama_index": "llamaindex", "open-webui/open-webui": "open_webui",
|
| 78 |
+
"huggingface/text-generation-inference": "tgi", "replicate/cog": "cog",
|
| 79 |
+
"stackblitz-labs/bolt.diy": "vite_react", "microsoft/sample-app-aoai-chatGPT": "flask",
|
| 80 |
+
"langgenius/dify": "dify", "infiniflow/ragflow": "ragflow", "Mintplex-Labs/anything-llm": "anythingllm",
|
| 81 |
+
"huggingface/chat-ui": "hf_chat_ui", "vercel/chatbot": "nextjs", "BerriAI/litellm": "litellm",
|
| 82 |
+
"vllm-project/vllm": "vllm", "mastra-ai/mastra": "mastra", "google/adk-python": "google_adk",
|
| 83 |
+
"crewAIInc/crewAI": "crewai", "agno-agi/agno": "agno", "skypilot-org/skypilot": "skypilot",
|
| 84 |
+
"Azure-Samples/rag-postgres-openai-python": "fastapi", "Azure-Samples/contoso-chat": "prompty"}
|
| 85 |
+
|
| 86 |
+
PROVIDERS = [
|
| 87 |
+
("azure_openai", rx(r"azure[ _-]?open ?ai|\baoai\b")), ("openai", rx(r"\bopen ?ai\b|\bgpt-?[345o]|chatgpt")),
|
| 88 |
+
("anthropic", rx(r"anthropic|\bclaude\b(?![- ]code)(?! made| told| deployed| concludes)")), ("google", rx(r"gemini|vertex ?ai|palm\b")),
|
| 89 |
+
("aws_bedrock", rx(r"bedrock")), ("local_ollama", rx(r"\bollama\b|llama\.cpp|lm studio|\bgguf\b")),
|
| 90 |
+
("huggingface", rx(r"hugging ?face|transformers|\bhf_|inference endpoint|\btgi\b")),
|
| 91 |
+
("mistral", rx(r"mistral")), ("deepseek", rx(r"deepseek")), ("groq", rx(r"\bgroq\b")),
|
| 92 |
+
("openrouter", rx(r"openrouter")), ("replicate", rx(r"replicate")), ("cohere", rx(r"cohere")),
|
| 93 |
+
]
|
| 94 |
+
REPO_PROVIDER = {"Azure-Samples/azure-search-openai-demo": "azure_openai",
|
| 95 |
+
"Azure-Samples/chat-with-your-data-solution-accelerator": "azure_openai",
|
| 96 |
+
"microsoft/sample-app-aoai-chatGPT": "azure_openai", "GoogleCloudPlatform/generative-ai": "google",
|
| 97 |
+
"GoogleCloudPlatform/agent-starter-pack": "google", "Azure-Samples/contoso-chat": "azure_openai",
|
| 98 |
+
"Azure-Samples/rag-postgres-openai-python": "azure_openai",
|
| 99 |
+
"ollama/ollama": "local_ollama", "huggingface/text-generation-inference": "huggingface"}
|
| 100 |
+
|
| 101 |
+
# credential words for secret_exposed; "token" alone is not one here (LLM tokens), it needs a qualifier
|
| 102 |
+
_CRED = (r"(secrets?\b|credentials?|api[_ -]?keys?|(access|auth|bearer|github|gh|hf|hugging ?face|personal access|oauth|refresh|session|jwt|api|secret)[_ -]?tokens?\b|"
|
| 103 |
+
r"passwords?|\.env\b|\.dev\.vars|private[_ -]?keys?|sensitive (files?|data|values?|info\w*))")
|
| 104 |
+
_SEC_ACT = r"(?<!avoid )(?<!not )(?<!never )(\bleak\w*|\bexpos(ed|es|ing)\b|(deployed|deploys|published|publishes|uploaded|uploads) |(in |as )?plain ?text)"
|
| 105 |
+
_SEC_ACT_AFTER = r"(\bleak\w*|\bexpos\w*|(in|as) plain ?text|publicly|made public|(is|are|was|were|now) public\b|(in|to|into) (a |the |my )?public (repo\w*|github|bundle|url))"
|
| 106 |
+
ERRORS = [ # error_type
|
| 107 |
+
("oom", rx(r"memory leak\w*|leak\w* memory|memory (usage|use|consumption) (keeps|continues|grows|growing|increas\w*)|ram growth|out of memory|\boom\b|oomkill|memory limit|heap out|killed.*memory|cuda out of memory|exceeded memory")),
|
| 108 |
+
("size_limit", rx(r"(volume|disk|storage).{0,25}(capacity|full|space|running low)|(size|bundle|function|slug|image|payload|body).{0,25}(limit|too large|exceed)|250 ?mb|50 ?mb|exceeds the maximum|too big|request entity too large|\b413\b")),
|
| 109 |
+
("timeout", rx(r"timed? ?out\b|timeout (error|exceeded|after)|\b504\b|deadline exceeded|maxduration|max duration|took too long|\d+ ?(s|sec|seconds|minutes) (limit|timeout)")),
|
| 110 |
+
("missing_env_var", rx(r"env(ironment)? ?var\w*.{0,40}(not|missing|undefined|empty)|(missing|undefined|not set|not found).{0,30}(api[_ ]?key|env|secret|token)|api[_ ]?key.{0,30}(missing|not (set|found|provided)|undefined|invalid|incorrect)|\.env\b")),
|
| 111 |
+
# credentials deployed/published by mistake: a credential word must sit next to the leak/expose/publish word
|
| 112 |
+
# ("memory leak", "exposed port", "context leaking" fall through to oom, port_binding, ...)
|
| 113 |
+
("secret_exposed", rx(_SEC_ACT + r"(?:(?!\. )[^\n]){0,30}" + _CRED + "|" + _CRED + r"(?:(?!\. )[^\n]){0,40}" + _SEC_ACT_AFTER)),
|
| 114 |
+
("iac_template_error", rx(r"invalidtemplate|template (deployment|validation)|bicep|terraform|arm template|deploymentfailed|resourcedeploymentfailure|failing invoking action .provision|error deploying infrastructure|cloudformation")),
|
| 115 |
+
("quota_billing", rx(r"quota|billing|payment|insufficient (funds|credits)|rate limit|\b429\b|credit card|\bcapacity\b|usage limit|limit of number")),
|
| 116 |
+
("permission_rbac", rx(r"permission denied|forbidden|\b403\b|unauthori[sz]ed|\b401\b|role assignment|rbac|authorizationfailed|not authorized|access denied")),
|
| 117 |
+
("port_binding", rx(r"\bport\b.{0,30}(\d{2,5}|bind|listen|expos)|\b(port|host) 0\.0\.0\.0|0\.0\.0\.0|failed to listen|listen(ing)? on port|health ?check|127\.0\.0\.1")),
|
| 118 |
+
("module_dependency", rx(r"module not found|cannot find module|modulenotfound|no module named|could not resolve|importerror|import error|dependency|peer dep|requirements\.txt|pip install|not compatible|incompatible|edge runtime")),
|
| 119 |
+
("build_failure", rx(r"build (fail|error)|failed to build|failed during stage .building|build script returned|npm run build|next build|exit code [1-9]|compil\w* (error|fail)|type (error|issue)s?|lockfile|frozen-lockfile")),
|
| 120 |
+
("cors_network", rx(r"\bcors\b|network error|econnrefused|connection refused|ssl|certificate|dns|websocket|\bwss?\b|proxy")),
|
| 121 |
+
("streaming", rx(r"stream(ing)?.{0,40}(not|doesn|broken|buffer|stuck|all at once|cut)|buffer(ed|ing)|(responses?|output) (are |is )?cut off")),
|
| 122 |
+
("not_found_routing", rx(r"\b404\b|page not found|routing|blank page|white (page|screen)|root[- ]path|sub ?path|base ?path|redirect")),
|
| 123 |
+
("server_error_5xx", rx(r"\b50[0-3]\b|internal ?server ?error|bad gateway|crash\w*|exited with|segmentation fault|application error|server-side exception")),
|
| 124 |
+
]
|
| 125 |
+
|
| 126 |
+
CATEGORY_OF_ERROR = {"secret_exposed": "secrets_api_keys", "iac_template_error": "provisioning_iac", "oom": "memory_model_size", "size_limit": "memory_model_size", "timeout": "timeouts",
|
| 127 |
+
"missing_env_var": "secrets_api_keys", "quota_billing": "costs_quota",
|
| 128 |
+
"permission_rbac": "auth_permissions", "port_binding": "networking",
|
| 129 |
+
"module_dependency": "runtime_dependencies", "build_failure": "build",
|
| 130 |
+
"cors_network": "networking", "streaming": "streaming",
|
| 131 |
+
"not_found_routing": "routing_config", "server_error_5xx": "other"}
|
| 132 |
+
CATEGORY_EXTRA = [
|
| 133 |
+
("gpu", rx(r"\bgpu|cuda|nvidia|a100|h100|\bt4\b|\bl4\b")),
|
| 134 |
+
("cold_start", rx(r"cold start|spin(s|ning)? (up|down)|sleep(ing)? (mode|after)|scale to zero|wake")),
|
| 135 |
+
("database", rx(r"database|postgres|mysql|sqlite|redis|mongo|vector (db|store)|cosmos|pgvector|chroma|pinecone|supabase")),
|
| 136 |
+
("auth_permissions", rx(r"\boauth|login|sign[- ]?in|\bsso\b|entra|easy auth|authentication")),
|
| 137 |
+
("costs_quota", rx(r"\bcosts?\b|pricing|expensive|\bbill(ed|ing)?\b|hobby plan|pro plan")),
|
| 138 |
+
]
|
| 139 |
+
|
| 140 |
+
# Topics added 2026-10-30, used ONLY when the rules above give "other" (existing topics always win).
|
| 141 |
+
# agent_behavior: a coding/deploy agent itself causes the problem. Needs an agent context (the claude-code
|
| 142 |
+
# repo, or the text names an agent); the subtype regex is tried on the title first, then the body.
|
| 143 |
+
AGENT_CTX = rx(r"\bclaude\b|\bthe (ai|agent|model|assistant)\b|coding agent|\bai agent|\bcodex\b|copilot|\bcursor\b|windsurf|auto[- ]?mode")
|
| 144 |
+
_DEP = r"(deploy\w*|push\w*|production|prod\b|merge\w*|release\w*)"
|
| 145 |
+
AGENT_ERRORS = [ # (error_type, regex); order = priority
|
| 146 |
+
("agent_wrong_target", rx(r"deploy\w* (\w+ )?to (the |a )?wrong|wrong (\w+ ){0,2}(project|environment|account|repo\w*|server|cluster|org\w*|vm|region|branch)\b")),
|
| 147 |
+
("agent_false_success", rx(r"false (completion|success)|(claim\w*|report\w*)[^.\n]{0,25}(unverified|untested) deploy|claim\w* (to (have )?)?deploy|"
|
| 148 |
+
r"report\w* (broken|failed) \w+ as (working|success\w*)|checks? pass\w* incorrectly|"
|
| 149 |
+
r"(claim\w*|report\w*|said|says|told)[^.\n]{0,40}(deploy\w*|production)[^.\n]{0,30}(work\w*|succe\w*|verified|live|fixed)[^.\n]{0,40}(but|when|while|although)|"
|
| 150 |
+
r"reported what a deploy|told [^.\n]{0,40}deploy[^.\n]{0,60}without (checking|verifying)")),
|
| 151 |
+
("agent_deployed_unasked", rx(_DEP + r"[^.\n]{0,40}(without (asking (for |me )?|any |explicit |user |prior |my )?(permission|confirmation|authori[sz]ation|approval|consent|being asked))|"
|
| 152 |
+
r"(autonomously|immediately|unilaterally) " + _DEP + r"|deploy\w*[^.\n]{0,20}(proceeded )?despite (explicit )?(non-deployment|instructions)|"
|
| 153 |
+
r"unintended deploy\w*|unreviewed deploy\w*|unasked deploy\w*|deploy\w* (proceeded|happened) despite")),
|
| 154 |
+
("agent_blocked_refused", rx(r"(auto[- ]?mode|classifier|safety (filter|flag|check)\w*|false positive)[^.\n]{0,60}(block\w*|den(y|ies|ied|ial)\w*|refus\w*|flag\w*)|"
|
| 155 |
+
r"(block\w*|refus\w*|den(y|ies|ied))[^.\n]{0,40}(" + _DEP[1:-1] + r")[^.\n]{0,40}(user[- ](authori[sz]ed|requested|approv\w*)|explicit\w*|valid|legitimate|approv\w*|authori[sz]\w*)|"
|
| 156 |
+
r"refus\w* [^.\n]{0,35}(merge|deploy)|false positive[^.\n]{0,40}(deploy|production)|flag\w*[^.\n]{0,40}blocks? (the )?deploy")),
|
| 157 |
+
]
|
| 158 |
+
# observability: monitoring/tracing/telemetry tools and terms (never bare "log"/"logs").
|
| 159 |
+
# The body regexes are stricter (bodies paste templates such as "Telemetry ID", and stack traces) and run on the
|
| 160 |
+
# cleaned first 1,500 characters of the body (no code blocks); OpenTelemetry and Application Insights count in the
|
| 161 |
+
# title only (in bodies they are mostly module paths in stack traces and Azure template resources).
|
| 162 |
+
OBSERVABILITY = rx(r"langfuse|langsmith (trac\w*|observab\w*|monitor\w*)|opentelemetry|\botel\b|\botlp\b|telemetry|\btracing\b|(?<!stack )\btraces\b|\btracer\b|dd-trace|"
|
| 163 |
+
r"\bsentry\b|prometheus|grafana|datadog|servicemonitor|observability|\bmonitoring\b|\bmetrics\b|helicone|\barize\b|new relic|application insights")
|
| 164 |
+
OBSERVABILITY_BODY = rx(r"langfuse|\botlp\b|\btracing\b(?! (the|this|it|back|down))|prometheus|grafana|datadog|\bsentry\b|servicemonitor|"
|
| 165 |
+
r"observability|helicone|\barize\b")
|
| 166 |
+
# performance_scaling: slowness, latency, throughput, concurrency, scaling out
|
| 167 |
+
PERFORMANCE = rx(r"\bslow(ly|ness|er|s)?\b|latency|throughput|concurren\w*|auto-?scal\w*|horizontal\w* scal\w*|scale (out|horizontally)|"
|
| 168 |
+
r"multi-?replicas?|\breplicas\b|single (api )?replica|in parallel|parallel (requests|agents|workers)|requests per second|tokens?/s\b|"
|
| 169 |
+
r"\[performance\]|performance (issue|problem|degrad\w*)|load balancing|high load|under load")
|
| 170 |
+
PERFORMANCE_BODY = rx(r"auto-?scal(ing|er)\b|horizontal\w* scal\w*|scale (out|horizontally)|multiple (replicas|workers)|multi-?replicas?|"
|
| 171 |
+
r"high(ly)? availab\w*|load balancing|throughput|concurrent (requests|users)|requests per second|tokens?/s\b|"
|
| 172 |
+
r"(very|extremely|too|really|painfully) slow|slow (responses?|startup|start|inference|performance)|high latency|latency (spikes?|increase\w*)")
|
| 173 |
+
|
| 174 |
+
|
| 175 |
+
def new_topic(repo, title, body):
|
| 176 |
+
"""(topic, agent error_type or None) for a thread whose topic is otherwise 'other', or (None, None).
|
| 177 |
+
Title first, then body (body = the first 3,000 raw characters, as for error_type)."""
|
| 178 |
+
agent_ok = repo == "anthropics/claude-code" or AGENT_CTX.search(title + "\n" + body)
|
| 179 |
+
e = agent_ok and first(AGENT_ERRORS, title)
|
| 180 |
+
if e:
|
| 181 |
+
return "agent_behavior", e
|
| 182 |
+
if OBSERVABILITY.search(title):
|
| 183 |
+
return "observability", None
|
| 184 |
+
if PERFORMANCE.search(title):
|
| 185 |
+
return "performance_scaling", None
|
| 186 |
+
e = agent_ok and first(AGENT_ERRORS, body)
|
| 187 |
+
if e:
|
| 188 |
+
return "agent_behavior", e
|
| 189 |
+
short = clean(body)[:1500]
|
| 190 |
+
if OBSERVABILITY_BODY.search(short):
|
| 191 |
+
return "observability", None
|
| 192 |
+
if PERFORMANCE_BODY.search(short):
|
| 193 |
+
return "performance_scaling", None
|
| 194 |
+
return None, None
|
| 195 |
+
|
| 196 |
+
|
| 197 |
+
AI_TOOLS = [("lovable", rx(r"lovable")), ("bolt", rx(r"\bbolt(\.new|\.diy)?\b")), ("v0", rx(r"v0\.dev|\bv0 (by vercel|app|generated)|(with|in|using|from) v0\b(?!\.)")),
|
| 198 |
+
("cursor", rx(r"\bcursor\b")), ("claude_code", rx(r"claude[- ]code")), ("replit", rx(r"replit")),
|
| 199 |
+
("copilot", rx(r"copilot")), ("windsurf", rx(r"windsurf")), ("dyad", rx(r"\bdyad\b")),
|
| 200 |
+
("codex", rx(r"\bcodex\b"))]
|
| 201 |
+
REPO_AI_TOOL = {"stackblitz-labs/bolt.diy": "bolt", "dyad-sh/dyad": "dyad", "anthropics/claude-code": "claude_code"}
|
| 202 |
+
|
| 203 |
+
REMOTE = rx(r"vercel|netlify|cloudflare|fly\.io|flyctl|railway|render|heroku|azure|azd|cloud ?run|gcp|google cloud|aws|lambda|"
|
| 204 |
+
r"sagemaker|ecs|kubernetes|k8s|spaces|streamlit cloud|production|prod\b|server|hosted|vps|digitalocean|runpod|modal|"
|
| 205 |
+
r"replicate|supabase|langgraph (platform|cloud)|deployed|deployment|serverless")
|
| 206 |
+
|
| 207 |
+
|
| 208 |
+
def clean(t):
|
| 209 |
+
t = re.sub(r"<!--.*?-->", " ", t or "", flags=re.S)
|
| 210 |
+
t = re.sub(r"```.*?```", " ", t, flags=re.S)
|
| 211 |
+
t = re.sub(r"`[^`]*`", " ", t)
|
| 212 |
+
t = re.sub(r"https?://\S+", "<url>", t)
|
| 213 |
+
t = re.sub(r"[\w.+-]+@[\w-]+\.[\w.]+", "<email>", t)
|
| 214 |
+
t = re.sub(r"@[\w-]+", "@<user>", t)
|
| 215 |
+
t = re.sub(r"^\s*(#+.*|[-*] \[[ x]\].*|\*\*[^*]+\*\*:?\s*)$", " ", t, flags=re.M)
|
| 216 |
+
t = re.sub(r"!\[[^\]]*\]\([^)]*\)", " ", t)
|
| 217 |
+
return re.sub(r"\s+", " ", t).strip()
|
| 218 |
+
|
| 219 |
+
|
| 220 |
+
def excerpt(body, n=30):
|
| 221 |
+
w = clean(body).split()
|
| 222 |
+
return " ".join(w[:n])
|
| 223 |
+
|
| 224 |
+
|
| 225 |
+
def first(rules, text):
|
| 226 |
+
for lab, r in rules:
|
| 227 |
+
if r.search(text):
|
| 228 |
+
return lab
|
| 229 |
+
return None
|
| 230 |
+
|
| 231 |
+
|
| 232 |
+
def multi(rules, text, k=3):
|
| 233 |
+
return ";".join([lab for lab, r in rules if r.search(text)][:k]) or None
|
| 234 |
+
|
| 235 |
+
|
| 236 |
+
# relevance rules learned from the pilot's manual judgement (see report.md)
|
| 237 |
+
NOT_DEPLOY = rx(r"deployment[_ ]?name|\bptu\b|serverless (api )?endpoint|bedrock marketplace|"
|
| 238 |
+
r"(model|llm|embedding)s? (deployed|hosted) (via|with|by|in|on) (vllm|xinference|azure|tgi|ollama)|"
|
| 239 |
+
r"(locally|local(ly)?[- ]deployed|my own) (model|llm|embedding)|deployed (embedding|chat|llm) model|"
|
| 240 |
+
r"use (a |my )?(remote|locally) deployed")
|
| 241 |
+
LOCAL = rx(r"\blocal(ly)? deploy\w*|\blocal(ly)? (docker|private|offline|intranet )?(docker |compose )?(deploy\w*|setup|install\w*)|docker local|docker desktop|\bwsl2?\b|truenas|langgraph dev\b|\bon my (mac|laptop|pc|machine)|desktop app|\.exe\b|"
|
| 242 |
+
r"windows 1[01]\b.{0,40}(server|local)|test(ing)? (it |your agent )?locally|errors when testing locally")
|
| 243 |
+
PITCH = rx(r"we('re| are) building|our (open-source )?platform|open-source platform for|happy to (help|contribute)|"
|
| 244 |
+
r"\bwithdrawn\b|marketplace where agents|we have implemented")
|
| 245 |
+
FEATURE = rx(r"^\s*(\[?feat(ure)?( request)?\]?[:\s]|add (support|a |an )|support (for )?|idea:|enhance|request:|"
|
| 246 |
+
r"allow |make .* (easier|friendly)|proposal)")
|
| 247 |
+
CLASSIFIER = rx(r"classifier|\baup\b|usage[- ]policy|safeguard|flag(s|ged)? .{0,30}(malicious|policy)|prompt[- ]injection")
|
| 248 |
+
REMOTE_T = rx(r"vercel|netlify|cloudflare|fly\.io|flyctl|railway|render\b|heroku|azure|azd|cloud ?run|gcp|google cloud|aws|"
|
| 249 |
+
r"lambda|sagemaker|ecs|eks|aks|gke|kubernetes|k8s|helm|spaces|streamlit cloud|production|prod\b|server|hosted|"
|
| 250 |
+
r"vps|digitalocean|runpod|modal|replicate|supabase|langgraph (platform|cloud)|reasoning engine|agent engine|"
|
| 251 |
+
r"inference endpoint|app service|container app|coolify|amplify|fargate|deploy\w*|serverless|edge")
|
| 252 |
+
BUILDER_SELF = rx(r"(deploy\w*|host\w*|run\w*|install\w*) (the |custom |costum |own )?(bolt|blot)(\.diy)?\b|"
|
| 253 |
+
r"\bbolt(\.diy)? (on|to|in) |bolt\.diy (is|on)|coolify|easypanel|docker|cloudflare runtime")
|
| 254 |
+
|
| 255 |
+
|
| 256 |
+
# v3 (after the full-build sample): a generic cloud/vendor name in the title is not enough
|
| 257 |
+
HOSTING_T = rx(r"vercel|netlify|railway|render\.com|onrender|\b(on|to) render\b(?!ing)(?! (the|a|an|it|html|markdown|text|images?)\b)|render (deploy\w*|blueprint|service)\b|"
|
| 258 |
+
r"heroku|fly\.io|flyctl|cloud ?run|app service|container apps?|"
|
| 259 |
+
r"aws lambda|\b(on|to|in|via) lambda\b|lambda (deploy\w*|layers?|cold start|timeouts?|function url|@edge|runtime)|"
|
| 260 |
+
r"ecs\b|fargate|\beks\b|\bgke\b|\baks\b|kubernetes|\bk8s?\b|\bk3s\b|helm|openshift|hugging ?face spaces?|hf spaces?|"
|
| 261 |
+
r"\b(on|to|in) (hf |hugging ?face |a |the |my )?spaces?\b(?! (between|in|and|before|after|at|around|instead|characters?|key))|\bspaces? (sdk|hardware|build|container|secrets|app|demo)|zerogpu|"
|
| 262 |
+
r"sagemaker (deploy\w*|endpoint deploy\w*)|deploy\w* .{0,40}sagemaker|\b(on|in|to) (an? |the )?(aws|azure|gcp|google cloud|ec2|sagemaker)\b|"
|
| 263 |
+
r"runpod|modal\.com|\bon modal\b|modal (deploy|serve|labs)|cloudformation|reverse proxy|nginx|traefik|ingress|cloudflared?[ -]tunnels?|\[serve\]|agent ?engine|reasoning engine|langgraph (platform|cloud)|"
|
| 264 |
+
r"langsmith deploy|mastra cloud|supabase (edge )?functions?|cloudflare (workers?|pages)|workers?\.dev|amplify|"
|
| 265 |
+
r"elastic beanstalk|app engine|digital ?ocean|coolify|docker[- ]compose|\bdocker\b|podman|\bvps\b|ec2|azd\b|streamlit (community )?cloud|"
|
| 266 |
+
r"inference endpoints?|posit connect|sky serve|skyserve|firebase hosting|firebase (app|functions)|"
|
| 267 |
+
r"(remote|cloud|linux|ubuntu|production|prod|company|shared|dedicated) (server|vm|instance|host|gpu)s?|\bon (a |the |my )?(remote )?(server|vm)\b")
|
| 268 |
+
DEPLOY_VERB_T = rx(r"\bdeploy\w*|\bhosting\b|\bhosted\b|self[- ]host\w*|\bredeploy\w*|\bprovision\w*|\bpublish\w*|go(ing)? live|"
|
| 269 |
+
r"\b(in|on|for|to) (the )?(production|prod)\b(?)|production (only|server|environment|build|deployment|mode|setup)|(prod|production) (env|build)\b|"
|
| 270 |
+
r"\bon (the )?server\b|\bazd up\b")
|
| 271 |
+
NOT_DEPLOY_SENSE = rx(r"router|load ?balanc\w* (between|across) (models|deployments)|model deployment(?! (to|on|in) )|deployment[_ ]?(name|id)s?\b|"
|
| 272 |
+
r"custom[- ]priced|cost map|\[usage\]|"
|
| 273 |
+
r"deploy\w* (the )?(new |latest )?version|after (deploying|upgrading) (to )?(v|version)|gpu deployment|deployment (thinking|defaults|settings)|"
|
| 274 |
+
r"(ios|macos|android) deployment|deployment target (too|for ios|version)|publish(ing)? (to )?(npm|pypi|the package|package)|published (package|version)|npm publish")
|
| 275 |
+
# Rule 7: a cloud's managed AI API used as a client, not an app deployed on a host
|
| 276 |
+
MANAGED_API_T = rx(r"vertex ?ai(?! agent ?engine)|bedrock|azure ?open ?ai|\baoai\b|ai foundry|azure ai (services|inference)|batch (prediction|api|inference|job)|"
|
| 277 |
+
r"gemini api|openai api|workers ai|ai gateway|sagemaker(?! deploy)|\bvercel_ai_gateway|\bsagemaker_|"
|
| 278 |
+
r"(serverless|provisioned|ptu|global|standard) (api |model )?(endpoint|deployment|throughput)")
|
| 279 |
+
APP_HOST_T = rx(r"vercel(?! ai gateway)(?!_ai)|netlify|railway|render\.com|heroku|fly\.io|cloud ?run|app service|container apps?|lambda\b|ecs\b|fargate|eks|gke|aks|"
|
| 280 |
+
r"kubernetes|k8s|helm|openshift|spaces?\b|(on|to|in) sagemaker|deploy\w* .{0,40}sagemaker|agent ?engine|reasoning engine|docker|vps|\bazd\b|app engine|cloudflare (workers?|pages)(?! ai)|"
|
| 281 |
+
r"deploy\w* (my |the |our |a |an )?(app|agent|server|proxy|litellm|chatbot|bot|application|site|frontend|backend|ui)|"
|
| 282 |
+
r"deploy\w* (a |the )?(model|llm|endpoint) (to|on|in) (azure ml|azure machine learning|sagemaker|vertex ai endpoint)|managed online endpoint")
|
| 283 |
+
# Rule 9: CI / tests of a framework's cloud integration
|
| 284 |
+
CI_T = rx(r"\bci\b|\[ci\]|^ci[:(]|github actions?|gh actions?|workflows? (fail\w*|run|broken)|\be2e\b|end[- ]to[- ]end tests?|integration tests?|unit tests?|"
|
| 285 |
+
r"smoke ?tests?|regression tests?|flaky|\btests? (fail\w*|for|of|against|on|broken)|^\[?tests?\]?[:(]|ci[- ]failure|release (pipeline|workflow|job)")
|
| 286 |
+
# Rule 6: app behaviour bugs; they need a link to the host (deploy, config, runtime) to stay in
|
| 287 |
+
APP_BUG_T = rx(r"\b(ui|button|display\w*|markdown|chat history|hallucinat\w*|infinite loop|loops? forever|repeat(s|ing|ed)? (itself|the|output|answer)|"
|
| 288 |
+
r"output (is )?(wrong|garbled|empty|incorrect)|answers?|citations?|i18n|translation|dark mode|theme|layout|css|font|style|"
|
| 289 |
+
r"tool[- ]call\w*|function[- ]call\w*|structured output|tokeni[sz]\w*|accuracy|quality|reasoning (text|content)|thinking (text|content)|"
|
| 290 |
+
r"cost tracking|spend|usage (data|tracking)|logs? (clutter|spam)|telemetry|metrics?|cach(e|ed|ing) tokens?|prompt_tokens)\b")
|
| 291 |
+
HOST_LINK = rx(r"\b(deploy\w* (fail\w*|error\w*|broken|stuck|crash\w*|times? out)|fail\w* (to|when|while|after|on|in) (deploy\w*|start\w*|boot|run(ning)? (on|in))|"
|
| 292 |
+
r"(cannot|can'?t|unable to|won'?t|doesn'?t|does not|not) (deploy|start|boot|connect|reach|access|load|run)|works? (fine )?locally|local(ly)? (works|fine)|"
|
| 293 |
+
r"but not (on|in|when|after)|only (on|in|when) (deploy\w*|production|prod|spaces|vercel|k8s|kubernetes|docker|server)|"
|
| 294 |
+
r"env\w*|config\w*|port|proxy|https?\b|ssl|tls|cors|domain|subpath|root[_ ]path|base[_ ]?path|timeout|timed out|memory|oom|ram|disk|volume|storage|"
|
| 295 |
+
r"startup|start(s|ing)?|crash\w*|restart\w*|container|image|helm|chart|ingress|install\w*|permission|forbidden|40[134]|50[0234]|auth\w*|login|oauth|"
|
| 296 |
+
r"secret|keys?|database|postgres|redis|network\w*|dns|websocket|stream\w*|cold start|scal\w*|replica\w*|pod|node|gpu|cuda|nccl|driver|build\w*|"
|
| 297 |
+
r"read[- ]only|non[- ]?root|filesystem|mount\w*|bucket|s3|gcs|blob|iframe|static files?|assets?|routing|redirect\w*|404|session)\b")
|
| 298 |
+
CC_NOISE = rx(r"browser pane|built-in browser|in-app browser|desktop|warp\b|oauth|connector|plugin|email|turnstile|challenge page|sub-?agents?|cowork|msix|"
|
| 299 |
+
r"auto-?updat\w*|managed[- ](settings|mcp|config)|deployment (mode|surface|store|mount)|third-party deployment|custom-3p|org(anization)?[- ]skill|"
|
| 300 |
+
r"skills? (deploy|publish)|marketplace|remote control|voice mode|install\.sh|binarydeployment|ssh remote|deployment target|ios|"
|
| 301 |
+
r"model deployed instead|driver deployment|deployed instead of|56-agent|deploy(ed|s)? .{0,20}(sub-?)?agents\b")
|
| 302 |
+
# Rule 4: an agent that deploys and hits its own limits (classifier, permission denied, false "it's deployed")
|
| 303 |
+
CC_DEPLOY = rx(r"deploy\w*|production|\bprod\b|publish\w*|go live|vercel|netlify|cloudflare (workers?|pages)|firebase|railway|supabase|heroku|"
|
| 304 |
+
r"fly\.io|cloud ?run|kubernetes|k8s|nginx|\bvps\b|\bserver\b|\bdomain\b|website|\bsite\b|\bhosting\b")
|
| 305 |
+
CC_LIMIT = rx(r"classifier|auto[- ]?mode|blocked|blocks|denied|deny|refus\w*|flag\w*|false (positive|completion)|claim\w*|reported .{0,40}(success|working|deployed)|"
|
| 306 |
+
r"without (asking|confirmation|permission|authori[sz]ation|warning)|unauthori[sz]ed|unrequested|unprompted|autonomous\w*|"
|
| 307 |
+
r"(deployed|published|uploaded) .{0,40}(secret|credential|\.env|\.dev\.vars|sensitive)|secret|credential|wrong (project|account|repo)|destroy\w*|took down|broke")
|
| 308 |
+
GENERIC_RE = filters.GENERIC_REPOS
|
| 309 |
+
SERVING_REPOS = {"vllm-project/vllm", "ollama/ollama", "huggingface/text-generation-inference", "skypilot-org/skypilot", "replicate/cog"}
|
| 310 |
+
AGENT_DEPLOY_REPOS = {"google/adk-python", "GoogleCloudPlatform/agent-starter-pack", "langchain-ai/langgraph", "mastra-ai/mastra", "crewAIInc/crewAI", "agno-agi/agno"}
|
| 311 |
+
|
| 312 |
+
|
| 313 |
+
def relevance(t, title, text, alltext):
|
| 314 |
+
"""Rules v4 (owner decisions 2026-10-28). Returns (yes|no, unsure_reason or '').
|
| 315 |
+
unsure_reason marks rows the keywords cannot settle; they are judged by the reviewer."""
|
| 316 |
+
g, repo = t["group"], t["repo"]
|
| 317 |
+
generic = g in ("host", "framework_generic") or repo in GENERIC_RE
|
| 318 |
+
tclean = re.sub(r"vercel ai sdk|ai sdk|sandbox-vercel|@vercel/\S+|vercel/ai\b|vercel[_ ]ai[_ ]gateway|ai gateway", " ", title, flags=I)
|
| 319 |
+
hosting, verb = bool(HOSTING_T.search(tclean)), bool(DEPLOY_VERB_T.search(tclean))
|
| 320 |
+
# rule 8: general frameworks and host CLIs need AI evidence in the thread itself (title, body, comments)
|
| 321 |
+
if generic and not filters.AI_EVIDENCE.search(re.sub(r"gradio|streamlit|vercel|next\.?js|netlify", " ", alltext, flags=I)):
|
| 322 |
+
return "no", ""
|
| 323 |
+
if PITCH.search(text[:1500]):
|
| 324 |
+
return "no", ""
|
| 325 |
+
if repo == "anthropics/claude-code":
|
| 326 |
+
if CC_NOISE.search(title) or NOT_DEPLOY_SENSE.search(title):
|
| 327 |
+
return "no", ""
|
| 328 |
+
if not CC_DEPLOY.search(title):
|
| 329 |
+
return "no", ""
|
| 330 |
+
# rule 4: the agent deploying (or claiming to) and hitting limits; rule 2: local docker build is not a host
|
| 331 |
+
if re.search(r"local `?docker build|locally", title, I):
|
| 332 |
+
return "no", ""
|
| 333 |
+
if CC_LIMIT.search(title) or re.search(r"deploy\w*", title, I):
|
| 334 |
+
return "yes", "" if CC_LIMIT.search(title) and re.search(r"deploy\w*|production|prod\b|publish", title, I) else "cc:deploy_context_unclear"
|
| 335 |
+
return "no", "cc:host_word_only"
|
| 336 |
+
if CLASSIFIER.search(title):
|
| 337 |
+
return "no", ""
|
| 338 |
+
if NOT_DEPLOY.search(text) and not re.search(r"deploy\w* (to|on|in) (vercel|netlify|azure|aws|gcp|cloud|render|railway|k8s|kubernetes|hugging)", title, I):
|
| 339 |
+
return "no", ""
|
| 340 |
+
remote = bool(HOSTING_T.search(re.sub(r"docker[- ]?(compose)?|podman", " ", tclean, flags=I)))
|
| 341 |
+
if LOCAL.search(title) and not remote or LOCAL.search(text[:2500]) and not hosting:
|
| 342 |
+
return "no", "" # rule 2: laptop, WSL, local GPU, "local docker deployment"
|
| 343 |
+
if not remote and re.search(r"\b(macos|mac (intel|m[1-4])|windows(?!\.net)|apple silicon|rtx \d{4}|jetson|raspberry|offline (device|machine)|local (gpu|machine|computer))\b", title, I) \
|
| 344 |
+
and not re.search(r"works? (fine )?(on|in) (the |my )?local|locally|\bbut\b|\b(adk|mastra|azd|az|gcloud|vercel|wrangler|langgraph|flowise) (deploy|build|up|webapp)\b|deploy\w* (to|on) (?!(the |my |a )?(mac|windows|jetson|raspberry))|agent_engine|gh actions", title, I):
|
| 345 |
+
return "no", "" # rule 2: a local machine named in the title
|
| 346 |
+
if not (g in ("framework", "sample", "builder") or filters.AI.search(text)):
|
| 347 |
+
return "no", ""
|
| 348 |
+
if not (hosting or verb):
|
| 349 |
+
return "no", "" # e.g. "Azure provider does not support ...", "AWS Bedrock ... structured output"
|
| 350 |
+
# rule 7: managed AI API as a client
|
| 351 |
+
own_host = g == "sample" or repo in SERVING_REPOS or repo in AGENT_DEPLOY_REPOS
|
| 352 |
+
if MANAGED_API_T.search(title) and not APP_HOST_T.search(tclean):
|
| 353 |
+
if own_host and (verb or re.search(r"\b(on|to|in) (gcp |google )?(vertex ai|sagemaker)\b", title, I)):
|
| 354 |
+
pass # rule 3 / sample apps: the repo's own app or model server deployed on that cloud
|
| 355 |
+
elif verb and re.search(r"deploy\w*", tclean, I):
|
| 356 |
+
return "yes", "managed_api:deploy_word"
|
| 357 |
+
else:
|
| 358 |
+
return "no", ""
|
| 359 |
+
if NOT_DEPLOY_SENSE.search(title) and not hosting:
|
| 360 |
+
return "no", ""
|
| 361 |
+
ci = bool(CI_T.search(title)) or any(re.search(r"ci[- ]failure|^ci$|area:ci|testing", l, I) for l in t["labels"])
|
| 362 |
+
if ci: # rule 9: CI/tests of a cloud integration are in; other CI is out
|
| 363 |
+
return ("yes", "") if hosting else ("no", "ci:no_cloud_named" if verb else "")
|
| 364 |
+
if repo == "BerriAI/litellm" and not (hosting or re.search(r"deploy\w* (litellm|the proxy|proxy)|docker|helm", title, I)):
|
| 365 |
+
return "no", "" # LiteLLM "deployment" = a model entry in the router
|
| 366 |
+
if repo in SERVING_REPOS and not hosting:
|
| 367 |
+
# rules 2+3: model serving counts on a remote host only; "deploy a model" often means a local run
|
| 368 |
+
if re.search(r"\b(server|production|prod|cluster|multi[- ]node|remote|cloud|container|docker|instance|api access|concurren\w*)\b", text[:2500], I):
|
| 369 |
+
return "yes", "serving:host_not_named"
|
| 370 |
+
return "no", ""
|
| 371 |
+
if g == "host" and re.search(r"agent", title, I) and not re.search(r"app|deploy\w* (my|a|the)", title, I):
|
| 372 |
+
return "no", "host:agent_tooling"
|
| 373 |
+
if not hosting and not re.search(r"deploy\w*|hosting|hosted|self[- ]host|provision|publish|go(ing)? live|on (the )?server|azd up", tclean, I):
|
| 374 |
+
return "yes", "production_only" # only "in production": a host is implied, the link may be weak
|
| 375 |
+
# rule 6: an app bug seen on a server, with no link to the deploy, config or host runtime
|
| 376 |
+
if APP_BUG_T.search(re.sub(r"chat-ui|open[- ]?webui|chatbot[- ]ui", " ", title, flags=I)) and not HOST_LINK.search(title):
|
| 377 |
+
return "yes", "app_bug:host_link_unclear"
|
| 378 |
+
return "yes", ""
|
| 379 |
+
|
| 380 |
+
|
| 381 |
+
# thread_type (rule 5): what the thread is
|
| 382 |
+
TT_PREFIX = [("pr_ci", rx(r"^\s*(\[(ci|test|tests|chore|build)\]|(ci|test|tests|chore)(\(\w+\))?:)")),
|
| 383 |
+
("docs", rx(r"^\s*(\[(docs?|documentation)\]|docs?(\(\w+\))?:|documentation:)")),
|
| 384 |
+
("feature", rx(r"^\s*(\[(feat(ure)?( request)?|enhancement|rfc|proposal|idea|request)\]|(feat|feature|feature request|enhancement|rfc|proposal|idea|request)(\(\w+\))?:)")),
|
| 385 |
+
("question", rx(r"^\s*(\[(question|usage|help|support|discussion)\]|(question|usage|help)(\(\w+\))?:)")),
|
| 386 |
+
("problem", rx(r"^\s*(\[(bug|bug report|installation|regression|model|refund)\]|bug(\(\w+\))?:)"))]
|
| 387 |
+
ERRORISH = rx(r"\b(errors?|fail\w*|broken|break(s|ing)?|crash\w*|not work\w*|doesn'?t work|does not work|isn'?t working|cannot|can'?t|unable|won'?t|stuck|"
|
| 388 |
+
r"timeouts?|timed out|exception|bug|wrong|incorrect|invalid|denied|refused|rejected|blocked|undefined|not found|hang(s|ing)?|freez\w*|"
|
| 389 |
+
r"[45]\d\d|missing|issue with|problem with|no longer|regression|leak\w*|ignor\w*|without (asking|confirmation|permission))\b")
|
| 390 |
+
FEATURE_T = rx(r"^\s*(add(ing)?|support(ing)?|allow|enable|provide|implement|introduce|make|expose|offer|create|let)\b|feature request|\bplease add\b|would be (nice|great|good|helpful)|"
|
| 391 |
+
r"\bsupport for\b|\bproposal\b|\bshould (support|allow|be able)|\bwish\b|\bability to\b|\boption (to|for)\b|\bconfigurable\b")
|
| 392 |
+
DOCS_T = rx(r"\bdocs?\b|documentation|readme|\bguide\b|tutorial|instructions?|\bexamples? (for|of|on)\b|recipe|cookbook|how-?to guide")
|
| 393 |
+
QUESTION_T = rx(r"^\s*(how|what|which|why|where|when|is|are|can|could|does|do|should|any)\b|\?\s*$|\bhow (to|do|can|should)\b|is it possible|is there a way|"
|
| 394 |
+
r"best (way|practice)|recommend\w*|guidance|help needed|need help|question")
|
| 395 |
+
FEATURE_L = rx(r"enhancement|feature|rfc|proposal|integration request|idea")
|
| 396 |
+
DOCS_L = rx(r"documentation|\bdocs?\b")
|
| 397 |
+
QUESTION_L = rx(r"question|support|usage|answered|help wanted(?!.)|discussion")
|
| 398 |
+
BUG_L = rx(r"\bbug\b|regression|crash|has repro|possible bug|🐞")
|
| 399 |
+
CI_L = rx(r"ci[- ]failure|^ci$|area:ci|\btests?\b|flaky")
|
| 400 |
+
|
| 401 |
+
|
| 402 |
+
def thread_type(t, title):
|
| 403 |
+
labels = [l.lower() for l in t["labels"]]
|
| 404 |
+
for lab, r in TT_PREFIX:
|
| 405 |
+
if r.search(title):
|
| 406 |
+
if lab == "docs" and re.search(r"(fail|error|broken|not working|outdated|wrong|incorrect)", title, I) and not DOCS_T.search(title):
|
| 407 |
+
return "problem"
|
| 408 |
+
return lab
|
| 409 |
+
if CI_T.search(title) or any(CI_L.search(l) for l in labels):
|
| 410 |
+
return "pr_ci"
|
| 411 |
+
err = bool(ERRORISH.search(title))
|
| 412 |
+
if DOCS_T.search(title) and not err:
|
| 413 |
+
return "docs"
|
| 414 |
+
if any(DOCS_L.search(l) for l in labels) and not any(BUG_L.search(l) for l in labels):
|
| 415 |
+
return "docs"
|
| 416 |
+
if any(FEATURE_L.search(l) for l in labels) and not err:
|
| 417 |
+
return "feature"
|
| 418 |
+
if FEATURE_T.search(title) and not err:
|
| 419 |
+
return "feature"
|
| 420 |
+
if err or any(BUG_L.search(l) for l in labels):
|
| 421 |
+
return "problem"
|
| 422 |
+
if QUESTION_T.search(title) or any(QUESTION_L.search(l) for l in labels):
|
| 423 |
+
return "question"
|
| 424 |
+
if ERRORISH.search(t.get("body", "")[:1500]):
|
| 425 |
+
return "problem"
|
| 426 |
+
return "other"
|
| 427 |
+
|
| 428 |
+
|
| 429 |
+
def code(t):
|
| 430 |
+
title, body = t["title"], t["body"]
|
| 431 |
+
others_c = [c for c in t["comments"] if not c["by_author"] and not c["is_bot"]]
|
| 432 |
+
coms = " ".join(c["body"] for c in t["comments"])
|
| 433 |
+
text = title + "\n" + body
|
| 434 |
+
repo, g = t["repo"], t["group"]
|
| 435 |
+
out = {}
|
| 436 |
+
alltext = text + "\n" + coms
|
| 437 |
+
out["kw_relevant"], out["kw_unsure"] = relevance(t, title, text, alltext)
|
| 438 |
+
out["kw_thread_type"] = thread_type(t, title)
|
| 439 |
+
builder_self = g == "builder" and repo == "stackblitz-labs/bolt.diy" and bool(BUILDER_SELF.search(text[:2000]))
|
| 440 |
+
# host: title first, then body, then repo default
|
| 441 |
+
out["kw_target_host"] = first(HOSTS, title) or first(HOSTS, body) or REPO_HOST.get(repo) or "unspecified"
|
| 442 |
+
# app type
|
| 443 |
+
data_cue = re.search(r"\brag\b|retriev|vector|embedding|knowledge base|ai search|cognitive ?search|own data|your data|index(es)? ", text, I)
|
| 444 |
+
if g == "builder" and not builder_self:
|
| 445 |
+
at = "ai_generated_web_app"
|
| 446 |
+
elif re.search(r"\bagents?\b|langgraph|crew|reasoning engine|agent engine|adk\b", text, I) or repo in ("langchain-ai/langgraph", "mastra-ai/mastra", "google/adk-python", "crewAIInc/crewAI", "agno-agi/agno", "GoogleCloudPlatform/agent-starter-pack"):
|
| 447 |
+
at = "agent"
|
| 448 |
+
elif repo in ("Azure-Samples/azure-search-openai-demo", "Azure-Samples/chat-with-your-data-solution-accelerator", "infiniflow/ragflow", "Mintplex-Labs/anything-llm", "Azure-Samples/rag-postgres-openai-python", "Azure-Samples/contoso-chat") or data_cue:
|
| 449 |
+
at = "rag"
|
| 450 |
+
elif repo in ("huggingface/text-generation-inference", "replicate/cog", "ollama/ollama", "vllm-project/vllm", "BerriAI/litellm", "skypilot-org/skypilot") or re.search(r"serv(e|ing) (the |a )?model|inference (server|endpoint)|model serving|sagemaker endpoint", text, I):
|
| 451 |
+
at = "model_serving"
|
| 452 |
+
elif repo in ("ChatGPTNextWeb/NextChat", "mckaywrigley/chatbot-ui", "huggingface/chat-ui", "vercel/chatbot", "Chainlit/chainlit", "open-webui/open-webui", "microsoft/sample-app-aoai-chatGPT") or re.search(r"chat ?bot|\bchat\b", text, I):
|
| 453 |
+
at = "chatbot"
|
| 454 |
+
else:
|
| 455 |
+
at = "other"
|
| 456 |
+
out["kw_app_type"] = at
|
| 457 |
+
fw_default = "remix" if builder_self else (None if g == "builder" else REPO_FRAMEWORK.get(repo))
|
| 458 |
+
out["kw_framework"] = fw_default or first(FRAMEWORKS, text) or "unspecified"
|
| 459 |
+
stack = [fw_default] + [lab for lab, r in FRAMEWORKS if r.search(text)]
|
| 460 |
+
out["kw_stack"] = ";".join(dict.fromkeys(s for s in stack if s)) or None
|
| 461 |
+
prov_text = text if repo != "anthropics/claude-code" else re.sub(r"claude|anthropic", " ", text, flags=I)
|
| 462 |
+
out["kw_model_provider"] = multi(PROVIDERS, prov_text) or REPO_PROVIDER.get(repo) or "unspecified"
|
| 463 |
+
# error type and category: title evidence first, then body
|
| 464 |
+
err_t, err_b = first(ERRORS, title), first(ERRORS, body[:3000])
|
| 465 |
+
extra_t = first(CATEGORY_EXTRA, title)
|
| 466 |
+
if err_t == "iac_template_error": # provisioning failures: prefer the inner cause when stated
|
| 467 |
+
inner = first([e for e in ERRORS if e[0] in ("quota_billing", "permission_rbac")], body[:3000])
|
| 468 |
+
err_t = inner or err_t
|
| 469 |
+
if err_t:
|
| 470 |
+
cat, conf = CATEGORY_OF_ERROR[err_t], "title"
|
| 471 |
+
elif extra_t:
|
| 472 |
+
cat, conf = extra_t, "title"
|
| 473 |
+
elif err_t:
|
| 474 |
+
cat, conf = CATEGORY_OF_ERROR[err_t], "title"
|
| 475 |
+
elif err_b:
|
| 476 |
+
cat, conf = CATEGORY_OF_ERROR[err_b], "body"
|
| 477 |
+
else:
|
| 478 |
+
cat, conf = "other", "none"
|
| 479 |
+
err = err_t or err_b or "other_unknown"
|
| 480 |
+
if cat == "other": # topics added 2026-10-30; never replace an existing topic
|
| 481 |
+
nt, agent_err = new_topic(repo, title, body[:3000])
|
| 482 |
+
if nt:
|
| 483 |
+
cat = nt
|
| 484 |
+
# the agent subtype replaces other_unknown, or a server_error_5xx found only in the body;
|
| 485 |
+
# a 5xx seen in the title is kept. coding_confidence is left as is: it describes the evidence
|
| 486 |
+
# for error_type and the 14 original topics, not for the three topics added 2026-10-30
|
| 487 |
+
if agent_err and (err == "other_unknown" or (err == "server_error_5xx" and conf == "body")):
|
| 488 |
+
err = agent_err
|
| 489 |
+
out["kw_error_type"] = err
|
| 490 |
+
out["kw_topic"] = cat
|
| 491 |
+
out["coding_confidence"] = conf
|
| 492 |
+
# AI-generated app
|
| 493 |
+
tool = first(AI_TOOLS, text) if re.search(r"(built|made|generated|created|wrote|written|vibe)[^.]{0,40}(with|by|using)|vibe.?cod", text, I) else None
|
| 494 |
+
tool = REPO_AI_TOOL.get(repo) or tool
|
| 495 |
+
if builder_self:
|
| 496 |
+
out["kw_ai_generated_app"], out["kw_ai_tool"] = "no", None
|
| 497 |
+
else:
|
| 498 |
+
out["kw_ai_generated_app"] = "yes" if tool else "unknown"
|
| 499 |
+
out["kw_ai_tool"] = tool
|
| 500 |
+
# thread outcome
|
| 501 |
+
maint = [c for c in others_c if c.get("assoc") in ("OWNER", "MEMBER", "COLLABORATOR")]
|
| 502 |
+
out["maintainer_replied"] = "yes" if maint else "no"
|
| 503 |
+
stale = any(re.search(r"stale|inactiv|auto-?clos", l, I) for l in t["labels"]) or any(
|
| 504 |
+
re.search(r"stale|inactiv|automatically closed|no activity|closing (this|due)", c["body"], I) for c in t["comments"] if c["is_bot"] or not c["by_author"])
|
| 505 |
+
fix_words = r"(it )?works now|that (worked|fixed)|solved|resolved|fixed it|found (the|a) (fix|solution|problem|issue)|the (problem|issue) was|turns out|was related to|i fixed it|managed to"
|
| 506 |
+
author_ok = any(c["by_author"] and re.search(fix_words + r"|thank(s| you).{0,40}(work|fix|help)", c["body"], I) for c in t["comments"]) \
|
| 507 |
+
or bool(re.search(r"i fixed it|update: .{0,40}(fixed|solved)", body, I))
|
| 508 |
+
dup = any(re.search(r"duplicate", l, I) for l in t["labels"]) or t.get("state_reason") == "duplicate"
|
| 509 |
+
if t["state"] == "open":
|
| 510 |
+
resolved = "yes" if author_ok else "no"
|
| 511 |
+
else:
|
| 512 |
+
resolved = "yes" if (t.get("state_reason") == "completed" and not stale) or author_ok else "no"
|
| 513 |
+
out["resolved"] = resolved
|
| 514 |
+
ocoms = " ".join(c["body"] for c in others_c)
|
| 515 |
+
if dup:
|
| 516 |
+
rt = "duplicate"
|
| 517 |
+
elif not others_c:
|
| 518 |
+
rt = "self_resolved" if author_ok else ("stale_closed" if stale and t["state"] == "closed" else "unanswered")
|
| 519 |
+
elif re.search(r"not (intended|designed|supported|possible|planned)|limitation|by design|won'?t fix|out of scope|isn'?t supported|is not an? (actual )?(bug|flowise error)", ocoms, I) and t["state"] == "closed":
|
| 520 |
+
rt = "platform_limit_no_fix" if not re.search(r"your (deployment )?environment|not a \w+ error", ocoms, I) else "user_error_config"
|
| 521 |
+
elif re.search(r"fixed in|released in|(upgrade|update) to (the latest|v?\d)|new (version|release)|latest (version|release)|beta release|now fixed|merged a fix|merging a fix", ocoms, I):
|
| 522 |
+
rt = "version_upgrade"
|
| 523 |
+
elif re.search(r"downgrad|pin(ned)? to|roll(ed)? back to", coms, I):
|
| 524 |
+
rt = "version_downgrade"
|
| 525 |
+
elif re.search(r"workaround|work around|in the meantime|temporar|stopgap", coms, I):
|
| 526 |
+
rt = "workaround"
|
| 527 |
+
elif re.search(r"my (bad|mistake|fault)|typo|you forgot|user error|misconfig|your (deployment )?environment", coms, I):
|
| 528 |
+
rt = "user_error_config"
|
| 529 |
+
elif author_ok and re.search(r"\bset\b|config|setting|env(ironment)? var|--host|0\.0\.0\.0|add(ing)? .{0,20}to (your|the)|change", coms, I):
|
| 530 |
+
rt = "config_fix"
|
| 531 |
+
elif author_ok:
|
| 532 |
+
rt = "self_resolved"
|
| 533 |
+
elif stale and t["state"] == "closed":
|
| 534 |
+
rt = "stale_closed"
|
| 535 |
+
elif t["state"] == "open":
|
| 536 |
+
rt = "open_unresolved"
|
| 537 |
+
else:
|
| 538 |
+
rt = "closed_unclear"
|
| 539 |
+
out["kw_resolution_type"] = rt
|
| 540 |
+
return out
|
| 541 |
+
|
| 542 |
+
|
| 543 |
+
# repos removed after the full-build review (see repos_full.json _doc); kept in the pilot file
|
| 544 |
+
DROPPED_REPOS = set()
|
| 545 |
+
|
| 546 |
+
|
| 547 |
+
def main():
|
| 548 |
+
src = sys.argv[1] if len(sys.argv) > 1 else os.path.join(ROOT, "raw", "threads.jsonl")
|
| 549 |
+
dst = sys.argv[2] if len(sys.argv) > 2 else os.path.join(ROOT, "raw", "coded_kw.jsonl")
|
| 550 |
+
if "--drop-removed" in sys.argv:
|
| 551 |
+
DROPPED_REPOS.update({"skypilot-org/skypilot", "replicate/cog"})
|
| 552 |
+
with open(dst, "w") as f:
|
| 553 |
+
for line in open(src):
|
| 554 |
+
t = json.loads(line)
|
| 555 |
+
if t["repo"] in DROPPED_REPOS:
|
| 556 |
+
continue
|
| 557 |
+
if t["repo"] in filters.GENERIC_REPOS and t["group"] == "framework":
|
| 558 |
+
t["group"] = "framework_generic" # rule 8: Gradio counts as a general framework
|
| 559 |
+
rec = {k: v for k, v in t.items() if k not in ("body", "comments")}
|
| 560 |
+
rec.update(code(t))
|
| 561 |
+
rec["excerpt_private"] = excerpt(t["body"])
|
| 562 |
+
f.write(json.dumps(rec) + "\n")
|
| 563 |
+
|
| 564 |
+
|
| 565 |
+
if __name__ == "__main__":
|
| 566 |
+
main()
|
scripts/collect.py
ADDED
|
@@ -0,0 +1,161 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Collect candidate GitHub issue threads about deploying AI apps.
|
| 3 |
+
|
| 4 |
+
Read-only GitHub REST API (search + issues + comments). Token from `gh auth token`.
|
| 5 |
+
Steps:
|
| 6 |
+
1. one search query per (repo, title term): `repo:R is:issue "term" in:title created:>=since`
|
| 7 |
+
2. local filters (see filters.py) -> keep_filter True/False
|
| 8 |
+
3. per-repo cap (best-match order) and global per-author cap
|
| 9 |
+
4. fetch issue + comments for kept threads, compute metrics
|
| 10 |
+
Outputs (in ../raw/, local only, never published):
|
| 11 |
+
search_hits.jsonl all search hits (title, url, repo, filter decision; no body)
|
| 12 |
+
threads.jsonl kept threads with body/comment text truncated (for coding only)
|
| 13 |
+
Logins are never written: only sha256(salt + login).
|
| 14 |
+
Usage: python3 -I scripts/collect.py [config.json] [out_dir] (defaults: scripts/repos.json, raw/)
|
| 15 |
+
"""
|
| 16 |
+
import json, os, re, subprocess, sys, time, hashlib, urllib.request, urllib.parse
|
| 17 |
+
from datetime import datetime
|
| 18 |
+
|
| 19 |
+
HERE = os.path.dirname(os.path.abspath(__file__))
|
| 20 |
+
sys.path.insert(0, HERE)
|
| 21 |
+
import filters # noqa: E402
|
| 22 |
+
|
| 23 |
+
ROOT = os.path.dirname(HERE)
|
| 24 |
+
CFG_PATH = sys.argv[1] if len(sys.argv) > 1 else os.path.join(HERE, "repos.json")
|
| 25 |
+
RAW = sys.argv[2] if len(sys.argv) > 2 else os.path.join(ROOT, "raw")
|
| 26 |
+
os.makedirs(RAW, exist_ok=True)
|
| 27 |
+
CFG = json.load(open(CFG_PATH))
|
| 28 |
+
SALT_PATH = os.path.join(ROOT, "raw", "salt.txt")
|
| 29 |
+
SALT = open(SALT_PATH).read().strip()
|
| 30 |
+
TOKEN = subprocess.run(["gh", "auth", "token"], capture_output=True, text=True).stdout.strip()
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
def ahash(login):
|
| 34 |
+
return hashlib.sha256((SALT + (login or "")).encode()).hexdigest()[:16]
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
def get(url, params=None):
|
| 38 |
+
if params:
|
| 39 |
+
url += "?" + urllib.parse.urlencode(params)
|
| 40 |
+
for attempt in range(5):
|
| 41 |
+
req = urllib.request.Request(url, headers={
|
| 42 |
+
"Authorization": f"Bearer {TOKEN}",
|
| 43 |
+
"Accept": "application/vnd.github+json",
|
| 44 |
+
"X-GitHub-Api-Version": "2022-11-28",
|
| 45 |
+
"User-Agent": "hal9-research-pilot"})
|
| 46 |
+
try:
|
| 47 |
+
with urllib.request.urlopen(req, timeout=60) as r:
|
| 48 |
+
return json.load(r)
|
| 49 |
+
except urllib.error.HTTPError as e:
|
| 50 |
+
if e.code in (403, 429):
|
| 51 |
+
time.sleep(30 * (attempt + 1)); continue
|
| 52 |
+
if e.code == 422:
|
| 53 |
+
print("422", url, file=sys.stderr); return None
|
| 54 |
+
raise
|
| 55 |
+
return None
|
| 56 |
+
|
| 57 |
+
|
| 58 |
+
def ts(s):
|
| 59 |
+
return datetime.fromisoformat(s.replace("Z", "+00:00")) if s else None
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
def hours(a, b):
|
| 63 |
+
return round((b - a).total_seconds() / 3600, 1) if a and b else None
|
| 64 |
+
|
| 65 |
+
|
| 66 |
+
def search():
|
| 67 |
+
hits = {}
|
| 68 |
+
for r in CFG["repos"]:
|
| 69 |
+
for term in r["title_terms"]:
|
| 70 |
+
q = f'repo:{r["repo"]} is:issue "{term}" in:title created:>={CFG["since"]}'
|
| 71 |
+
items, total = [], None
|
| 72 |
+
for page in range(1, CFG.get("pages", 1) + 1):
|
| 73 |
+
res = get("https://api.github.com/search/issues", {"q": q, "per_page": 100, "page": page})
|
| 74 |
+
time.sleep(2.2) # search API: 30 requests/minute
|
| 75 |
+
if not res:
|
| 76 |
+
break
|
| 77 |
+
total = res.get("total_count")
|
| 78 |
+
items += res.get("items", [])
|
| 79 |
+
if len(items) >= min(total, 1000) or not res.get("items"):
|
| 80 |
+
break
|
| 81 |
+
for rank, it in enumerate(items):
|
| 82 |
+
if "pull_request" in it:
|
| 83 |
+
continue
|
| 84 |
+
u = it["html_url"]
|
| 85 |
+
h = hits.setdefault(u, {"url": u, "repo": r["repo"], "group": r["group"],
|
| 86 |
+
"title": it["title"], "created_at": it["created_at"],
|
| 87 |
+
"author_hash": ahash(it["user"]["login"]),
|
| 88 |
+
"body": (it.get("body") or "")[:6000],
|
| 89 |
+
"terms": [], "best_rank": rank, "_item": it})
|
| 90 |
+
h["terms"].append(term)
|
| 91 |
+
h["best_rank"] = min(h["best_rank"], rank)
|
| 92 |
+
print(f'{r["repo"]:50s} {term:20s} {total}', file=sys.stderr)
|
| 93 |
+
return list(hits.values())
|
| 94 |
+
|
| 95 |
+
|
| 96 |
+
def main():
|
| 97 |
+
hits = search()
|
| 98 |
+
for h in hits:
|
| 99 |
+
h["keep_filter"], h["filter_reason"] = filters.keep(h)
|
| 100 |
+
# caps: per repo by best rank (more matching terms first), then per author globally
|
| 101 |
+
kept, per_author = [], {}
|
| 102 |
+
by_repo = {}
|
| 103 |
+
for h in hits:
|
| 104 |
+
if h["keep_filter"]:
|
| 105 |
+
by_repo.setdefault(h["repo"], []).append(h)
|
| 106 |
+
repo_cap = {r["repo"]: r["cap"] for r in CFG["repos"] if "cap" in r}
|
| 107 |
+
for repo, lst in by_repo.items():
|
| 108 |
+
lst.sort(key=lambda h: (-len(h["terms"]), h["best_rank"]))
|
| 109 |
+
n = 0
|
| 110 |
+
for h in lst:
|
| 111 |
+
if n >= repo_cap.get(repo, CFG["per_repo_cap"]):
|
| 112 |
+
h["cap"] = "repo_cap"; continue
|
| 113 |
+
if per_author.get(h["author_hash"], 0) >= CFG["per_author_cap"]:
|
| 114 |
+
h["cap"] = "author_cap"; continue
|
| 115 |
+
per_author[h["author_hash"]] = per_author.get(h["author_hash"], 0) + 1
|
| 116 |
+
h["cap"] = "kept"; n += 1
|
| 117 |
+
kept.append(h)
|
| 118 |
+
with open(os.path.join(RAW, "search_hits.jsonl"), "w") as f:
|
| 119 |
+
for h in hits:
|
| 120 |
+
f.write(json.dumps({k: v for k, v in h.items() if k not in ("body", "_item")}) + "\n")
|
| 121 |
+
print(f"hits={len(hits)} pass_filter={sum(h['keep_filter'] for h in hits)} kept={len(kept)}", file=sys.stderr)
|
| 122 |
+
|
| 123 |
+
tpath = os.path.join(RAW, "threads.jsonl")
|
| 124 |
+
done = set()
|
| 125 |
+
if os.path.exists(tpath): # resume: skip threads already fetched
|
| 126 |
+
done = {json.loads(l)["url"] for l in open(tpath)}
|
| 127 |
+
out = open(tpath, "a")
|
| 128 |
+
for h in kept:
|
| 129 |
+
if h["url"] in done:
|
| 130 |
+
continue
|
| 131 |
+
owner_repo, num = h["repo"], h["url"].rsplit("/", 1)[1]
|
| 132 |
+
api = f"https://api.github.com/repos/{owner_repo}/issues/{num}"
|
| 133 |
+
iss = h["_item"] # search items carry state, state_reason, reactions, labels
|
| 134 |
+
coms = (get(api + "/comments", {"per_page": 100}) or []) if iss["comments"] else []
|
| 135 |
+
author = iss["user"]["login"]
|
| 136 |
+
created = ts(iss["created_at"])
|
| 137 |
+
first_other = next((c for c in coms if c["user"]["login"] != author
|
| 138 |
+
and not c["user"]["login"].endswith("[bot]")
|
| 139 |
+
and c["user"].get("type") != "Bot"), None)
|
| 140 |
+
rec = {
|
| 141 |
+
"source": "github_issues", "repo": owner_repo, "group": h["group"],
|
| 142 |
+
"url": iss["html_url"], "created_at": iss["created_at"], "title": iss["title"],
|
| 143 |
+
"n_comments": iss["comments"], "n_reactions": iss.get("reactions", {}).get("total_count", 0),
|
| 144 |
+
"state": iss["state"], "state_reason": iss.get("state_reason"),
|
| 145 |
+
"closed_at": iss.get("closed_at"),
|
| 146 |
+
"labels": [l["name"] for l in iss.get("labels", [])],
|
| 147 |
+
"time_to_first_response_h": hours(created, ts(first_other["created_at"])) if first_other else None,
|
| 148 |
+
"time_to_close_h": hours(created, ts(iss.get("closed_at"))),
|
| 149 |
+
"author_hash": ahash(author), "terms": h["terms"],
|
| 150 |
+
"body": (iss.get("body") or "")[:6000],
|
| 151 |
+
"comments": [{"by_author": c["user"]["login"] == author,
|
| 152 |
+
"is_bot": c["user"]["login"].endswith("[bot]") or c["user"].get("type") == "Bot",
|
| 153 |
+
"assoc": c.get("author_association"),
|
| 154 |
+
"body": (c.get("body") or "")[:1500]} for c in coms],
|
| 155 |
+
}
|
| 156 |
+
out.write(json.dumps(rec) + "\n"); out.flush()
|
| 157 |
+
out.close()
|
| 158 |
+
|
| 159 |
+
|
| 160 |
+
if __name__ == "__main__":
|
| 161 |
+
main()
|
scripts/filters.py
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Keyword rules shared by collect.py (relevance filter) and code.py (coding)."""
|
| 2 |
+
import re
|
| 3 |
+
|
| 4 |
+
I = re.I
|
| 5 |
+
DEPLOY_T = re.compile(r"\b(deploy\w*|hosting|hosted|production|prod\b|publish\w*|self-host\w*|"
|
| 6 |
+
r"cloud ?run|app service|azd|provision\w*|lambda|serverless|edge (function|runtime)|"
|
| 7 |
+
r"vercel|netlify|railway|render\.com|on render|fly\.io|heroku|cloudflare|"
|
| 8 |
+
r"kubernetes|k8s|helm|sagemaker|runpod|modal|spaces?\b|aws|azure|gcp|ecs|eks|gke|docker)", I)
|
| 9 |
+
AI = re.compile(r"\b(ai|llm\w*|openai|gpt[-\w]*|chatgpt|langchain\w*|langgraph|llama\w*|ollama|gradio|"
|
| 10 |
+
r"hugging ?face|transformers|model|embedding\w*|rag|chatbot|anthropic|claude|gemini|"
|
| 11 |
+
r"vertex|bedrock|mistral|vllm|tgi|inference|agent\w*|ai[- ]sdk|ai gateway|mcp|whisper|"
|
| 12 |
+
r"stable diffusion|torch|pytorch|tensorflow|gpu|cuda)\b", I)
|
| 13 |
+
# known false-positive senses of "deploy"/"deployment"
|
| 14 |
+
FALSE_SENSE = re.compile(r"(deployment[_ ]?name|azure_openai_deployment|model deployment(?! to)|"
|
| 15 |
+
r"deployment_id|\bdeployment\b of (gpt|the model)|azure ?openai deployment)", I)
|
| 16 |
+
# Rule 8 (owner decision 2026-10-28): in general-purpose frameworks (Streamlit, Gradio, Next.js, host CLIs)
|
| 17 |
+
# the thread itself must show an AI app; the framework's name is not evidence.
|
| 18 |
+
AI_EVIDENCE = re.compile(r"\b(ai|a\.i\.|llms?|large language|openai|gpt[-\w]*|chatgpt|langchain\w*|langgraph|llama\w*|ollama|"
|
| 19 |
+
r"hugging ?face|transformers|(ml|ai|llm|language|diffusion|whisper|vision|embedding|chat) models?|"
|
| 20 |
+
r"models? (inference|weights|loading|card|repo)|(load|run|serve|host)(s|ing)? (the |a |my )?models?|"
|
| 21 |
+
r"embedding\w*|rag|chat ?bot\w*|anthropic|claude|gemini|vertex ai|bedrock|mistral|qwen|deepseek|"
|
| 22 |
+
r"vllm|tgi|inference|agents?|ai[- ]sdk|mcp|whisper|stable diffusion|diffusers|automatic1111|comfyui|sdxl|lora|flux|text-to-\\w+|speech|tts|asr|ocr|torch|pytorch|tensorflow|"
|
| 23 |
+
r"cuda|gpu|machine learning|deep learning|zerogpu|prompt|tokens?|cursor|copilot|lovable|bolt|v0)\b", I)
|
| 24 |
+
GENERIC_REPOS = {"streamlit/streamlit", "gradio-app/gradio", "vercel/next.js", "vercel/vercel", "netlify/cli",
|
| 25 |
+
"superfly/flyctl", "railwayapp/cli"}
|
| 26 |
+
LOCAL_ONLY = re.compile(r"\b(localhost|local(ly)? (deploy|install|run)|on my (mac|laptop|pc)|windows 1[01])\b", I)
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def keep(h):
|
| 30 |
+
title, body, g = h["title"], h.get("body", ""), h["group"]
|
| 31 |
+
text = title + "\n" + body
|
| 32 |
+
if g == "host":
|
| 33 |
+
if AI.search(title):
|
| 34 |
+
return True, "host:ai_in_title"
|
| 35 |
+
if AI.search(body) and DEPLOY_T.search(title):
|
| 36 |
+
return True, "host:ai_in_body+deploy_title"
|
| 37 |
+
return False, "host:no_ai"
|
| 38 |
+
if not DEPLOY_T.search(title):
|
| 39 |
+
return False, "no_deploy_term_in_title"
|
| 40 |
+
if g == "framework_generic" and not AI_EVIDENCE.search(text):
|
| 41 |
+
return False, "generic_framework:no_ai"
|
| 42 |
+
if FALSE_SENSE.search(title) and not re.search(r"deploy(ed|ing)? (to|on)\b", title, I):
|
| 43 |
+
return False, "false_sense:model_deployment_name"
|
| 44 |
+
return True, f"{g}:deploy_title"
|
scripts/refresh.sh
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env bash
|
| 2 |
+
# Monthly refresh (local only; nothing is uploaded or published by this script).
|
| 3 |
+
# Re-collects every thread (states, comments and counts change), re-codes with keywords,
|
| 4 |
+
# re-applies the manual judgements (keyed by url), and rebuilds data/threads.csv.
|
| 5 |
+
# Needs: gh CLI logged in (read-only API use), python3. Runtime about 1-2 h, unattended.
|
| 6 |
+
# Usage: scripts/refresh.sh (from the dataset folder)
|
| 7 |
+
set -euo pipefail
|
| 8 |
+
cd "$(dirname "$0")/.."
|
| 9 |
+
month=$(date +%Y-%m)
|
| 10 |
+
out="raw_full_$month"
|
| 11 |
+
mkdir -p "$out"
|
| 12 |
+
python3 -I scripts/collect.py scripts/repos_full.json "$out" 2> "$out/collect.log"
|
| 13 |
+
python3 -I scripts/code.py "$out/threads.jsonl" "$out/coded_kw.jsonl" --drop-removed
|
| 14 |
+
python3 -I scripts/build_dataset.py "$out/coded_kw.jsonl" data/threads.csv private/verification.csv \
|
| 15 |
+
manual/judgement.csv manual/judgement_full.csv manual/judgement_validation.csv \
|
| 16 |
+
manual/judgement_v4.csv manual/judgement_v4_precision.csv
|
| 17 |
+
python3 -I scripts/stats.py data/threads.csv > "private/stats_$month.txt"
|
| 18 |
+
# Rows the keywords cannot settle have a non-empty kw_unsure in private/verification.csv: judge them too
|
| 19 |
+
# (relevance yes/no + thread_type, rules in decisions-2026-10-28.md) and append them to manual/judgement_v4.csv.
|
| 20 |
+
# Human step (about 1-2 h a month): judge a random sample of 50 new rows for relevance + codes
|
| 21 |
+
# python3 -I scripts/sample_for_review.py "$out/coded_kw.jsonl" 50 > private/review_$month.txt
|
| 22 |
+
# then append the judgements to manual/judgement_full.csv (codes) / judgement_v4.csv (relevance, thread_type)
|
| 23 |
+
# and re-run the build step above. Mark who judged in the judged_by column (person or agent).
|
| 24 |
+
echo "done: data/threads.csv rebuilt from $out"
|
scripts/repos_full.json
ADDED
|
@@ -0,0 +1,995 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_doc": "Full build. Host CLI repos, langchain (py) and llama_index dropped: pilot precision 0-21%. One query per (repo, title term), up to 3 pages each. After the full-build review, skypilot (GPU jobs, not app deploys: 1/5 relevant) and replicate/cog (2 threads) were removed too.",
|
| 3 |
+
"since": "2024-01-01",
|
| 4 |
+
"pages": 3,
|
| 5 |
+
"per_repo_cap": 250,
|
| 6 |
+
"per_author_cap": 2,
|
| 7 |
+
"repos": [
|
| 8 |
+
{
|
| 9 |
+
"repo": "streamlit/streamlit",
|
| 10 |
+
"group": "framework_generic",
|
| 11 |
+
"title_terms": [
|
| 12 |
+
"deploy",
|
| 13 |
+
"deployed",
|
| 14 |
+
"cloud run",
|
| 15 |
+
"heroku",
|
| 16 |
+
"community cloud",
|
| 17 |
+
"production"
|
| 18 |
+
]
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"repo": "gradio-app/gradio",
|
| 22 |
+
"group": "framework_generic",
|
| 23 |
+
"title_terms": [
|
| 24 |
+
"deploy",
|
| 25 |
+
"deployed",
|
| 26 |
+
"deployment",
|
| 27 |
+
"production",
|
| 28 |
+
"hosting",
|
| 29 |
+
"vercel",
|
| 30 |
+
"netlify",
|
| 31 |
+
"cloudflare",
|
| 32 |
+
"railway",
|
| 33 |
+
"render",
|
| 34 |
+
"heroku",
|
| 35 |
+
"fly.io",
|
| 36 |
+
"azure",
|
| 37 |
+
"aws",
|
| 38 |
+
"lambda",
|
| 39 |
+
"cloud run",
|
| 40 |
+
"gcp",
|
| 41 |
+
"kubernetes",
|
| 42 |
+
"k8s",
|
| 43 |
+
"sagemaker",
|
| 44 |
+
"app service",
|
| 45 |
+
"spaces",
|
| 46 |
+
"inference endpoint",
|
| 47 |
+
"ecs",
|
| 48 |
+
"supabase",
|
| 49 |
+
"serverless"
|
| 50 |
+
]
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"repo": "Chainlit/chainlit",
|
| 54 |
+
"group": "framework",
|
| 55 |
+
"title_terms": [
|
| 56 |
+
"deploy",
|
| 57 |
+
"deployed",
|
| 58 |
+
"deployment",
|
| 59 |
+
"production",
|
| 60 |
+
"hosting",
|
| 61 |
+
"vercel",
|
| 62 |
+
"netlify",
|
| 63 |
+
"cloudflare",
|
| 64 |
+
"railway",
|
| 65 |
+
"render",
|
| 66 |
+
"heroku",
|
| 67 |
+
"fly.io",
|
| 68 |
+
"azure",
|
| 69 |
+
"aws",
|
| 70 |
+
"lambda",
|
| 71 |
+
"cloud run",
|
| 72 |
+
"gcp",
|
| 73 |
+
"kubernetes",
|
| 74 |
+
"k8s",
|
| 75 |
+
"sagemaker",
|
| 76 |
+
"app service",
|
| 77 |
+
"spaces",
|
| 78 |
+
"inference endpoint",
|
| 79 |
+
"ecs",
|
| 80 |
+
"supabase",
|
| 81 |
+
"serverless"
|
| 82 |
+
]
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"repo": "langchain-ai/langchainjs",
|
| 86 |
+
"group": "framework",
|
| 87 |
+
"title_terms": [
|
| 88 |
+
"deploy",
|
| 89 |
+
"deployed",
|
| 90 |
+
"vercel",
|
| 91 |
+
"edge",
|
| 92 |
+
"lambda",
|
| 93 |
+
"cloudflare",
|
| 94 |
+
"netlify",
|
| 95 |
+
"production"
|
| 96 |
+
]
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"repo": "langchain-ai/langgraph",
|
| 100 |
+
"group": "framework",
|
| 101 |
+
"title_terms": [
|
| 102 |
+
"deploy",
|
| 103 |
+
"deployed",
|
| 104 |
+
"deployment",
|
| 105 |
+
"production",
|
| 106 |
+
"hosting",
|
| 107 |
+
"vercel",
|
| 108 |
+
"netlify",
|
| 109 |
+
"cloudflare",
|
| 110 |
+
"railway",
|
| 111 |
+
"render",
|
| 112 |
+
"heroku",
|
| 113 |
+
"fly.io",
|
| 114 |
+
"azure",
|
| 115 |
+
"aws",
|
| 116 |
+
"lambda",
|
| 117 |
+
"cloud run",
|
| 118 |
+
"gcp",
|
| 119 |
+
"kubernetes",
|
| 120 |
+
"k8s",
|
| 121 |
+
"sagemaker",
|
| 122 |
+
"app service",
|
| 123 |
+
"spaces",
|
| 124 |
+
"inference endpoint",
|
| 125 |
+
"ecs",
|
| 126 |
+
"supabase",
|
| 127 |
+
"serverless"
|
| 128 |
+
]
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"repo": "vercel/ai",
|
| 132 |
+
"group": "framework",
|
| 133 |
+
"title_terms": [
|
| 134 |
+
"deploy",
|
| 135 |
+
"deployed",
|
| 136 |
+
"deployment",
|
| 137 |
+
"production",
|
| 138 |
+
"hosting",
|
| 139 |
+
"vercel",
|
| 140 |
+
"netlify",
|
| 141 |
+
"cloudflare",
|
| 142 |
+
"railway",
|
| 143 |
+
"render",
|
| 144 |
+
"heroku",
|
| 145 |
+
"fly.io",
|
| 146 |
+
"azure",
|
| 147 |
+
"aws",
|
| 148 |
+
"lambda",
|
| 149 |
+
"cloud run",
|
| 150 |
+
"gcp",
|
| 151 |
+
"kubernetes",
|
| 152 |
+
"k8s",
|
| 153 |
+
"sagemaker",
|
| 154 |
+
"app service",
|
| 155 |
+
"spaces",
|
| 156 |
+
"inference endpoint",
|
| 157 |
+
"ecs",
|
| 158 |
+
"supabase",
|
| 159 |
+
"serverless"
|
| 160 |
+
]
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"repo": "ollama/ollama",
|
| 164 |
+
"group": "framework",
|
| 165 |
+
"title_terms": [
|
| 166 |
+
"deploy",
|
| 167 |
+
"deployed",
|
| 168 |
+
"deployment",
|
| 169 |
+
"production",
|
| 170 |
+
"hosting",
|
| 171 |
+
"vercel",
|
| 172 |
+
"netlify",
|
| 173 |
+
"cloudflare",
|
| 174 |
+
"railway",
|
| 175 |
+
"render",
|
| 176 |
+
"heroku",
|
| 177 |
+
"fly.io",
|
| 178 |
+
"azure",
|
| 179 |
+
"aws",
|
| 180 |
+
"lambda",
|
| 181 |
+
"cloud run",
|
| 182 |
+
"gcp",
|
| 183 |
+
"kubernetes",
|
| 184 |
+
"k8s",
|
| 185 |
+
"sagemaker",
|
| 186 |
+
"app service",
|
| 187 |
+
"spaces",
|
| 188 |
+
"inference endpoint",
|
| 189 |
+
"ecs",
|
| 190 |
+
"supabase",
|
| 191 |
+
"serverless"
|
| 192 |
+
]
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"repo": "ChatGPTNextWeb/NextChat",
|
| 196 |
+
"group": "framework",
|
| 197 |
+
"title_terms": [
|
| 198 |
+
"deploy",
|
| 199 |
+
"deployed",
|
| 200 |
+
"deployment",
|
| 201 |
+
"production",
|
| 202 |
+
"hosting",
|
| 203 |
+
"vercel",
|
| 204 |
+
"netlify",
|
| 205 |
+
"cloudflare",
|
| 206 |
+
"railway",
|
| 207 |
+
"render",
|
| 208 |
+
"heroku",
|
| 209 |
+
"fly.io",
|
| 210 |
+
"azure",
|
| 211 |
+
"aws",
|
| 212 |
+
"lambda",
|
| 213 |
+
"cloud run",
|
| 214 |
+
"gcp",
|
| 215 |
+
"kubernetes",
|
| 216 |
+
"k8s",
|
| 217 |
+
"sagemaker",
|
| 218 |
+
"app service",
|
| 219 |
+
"spaces",
|
| 220 |
+
"inference endpoint",
|
| 221 |
+
"ecs",
|
| 222 |
+
"supabase",
|
| 223 |
+
"serverless"
|
| 224 |
+
]
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"repo": "mckaywrigley/chatbot-ui",
|
| 228 |
+
"group": "framework",
|
| 229 |
+
"title_terms": [
|
| 230 |
+
"deploy",
|
| 231 |
+
"deployed",
|
| 232 |
+
"deployment",
|
| 233 |
+
"production",
|
| 234 |
+
"hosting",
|
| 235 |
+
"vercel",
|
| 236 |
+
"netlify",
|
| 237 |
+
"cloudflare",
|
| 238 |
+
"railway",
|
| 239 |
+
"render",
|
| 240 |
+
"heroku",
|
| 241 |
+
"fly.io",
|
| 242 |
+
"azure",
|
| 243 |
+
"aws",
|
| 244 |
+
"lambda",
|
| 245 |
+
"cloud run",
|
| 246 |
+
"gcp",
|
| 247 |
+
"kubernetes",
|
| 248 |
+
"k8s",
|
| 249 |
+
"sagemaker",
|
| 250 |
+
"app service",
|
| 251 |
+
"spaces",
|
| 252 |
+
"inference endpoint",
|
| 253 |
+
"ecs",
|
| 254 |
+
"supabase",
|
| 255 |
+
"serverless"
|
| 256 |
+
]
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"repo": "FlowiseAI/Flowise",
|
| 260 |
+
"group": "framework",
|
| 261 |
+
"title_terms": [
|
| 262 |
+
"deploy",
|
| 263 |
+
"deployed",
|
| 264 |
+
"deployment",
|
| 265 |
+
"production",
|
| 266 |
+
"hosting",
|
| 267 |
+
"vercel",
|
| 268 |
+
"netlify",
|
| 269 |
+
"cloudflare",
|
| 270 |
+
"railway",
|
| 271 |
+
"render",
|
| 272 |
+
"heroku",
|
| 273 |
+
"fly.io",
|
| 274 |
+
"azure",
|
| 275 |
+
"aws",
|
| 276 |
+
"lambda",
|
| 277 |
+
"cloud run",
|
| 278 |
+
"gcp",
|
| 279 |
+
"kubernetes",
|
| 280 |
+
"k8s",
|
| 281 |
+
"sagemaker",
|
| 282 |
+
"app service",
|
| 283 |
+
"spaces",
|
| 284 |
+
"inference endpoint",
|
| 285 |
+
"ecs",
|
| 286 |
+
"supabase",
|
| 287 |
+
"serverless"
|
| 288 |
+
]
|
| 289 |
+
},
|
| 290 |
+
{
|
| 291 |
+
"repo": "langflow-ai/langflow",
|
| 292 |
+
"group": "framework",
|
| 293 |
+
"title_terms": [
|
| 294 |
+
"deploy",
|
| 295 |
+
"deployed",
|
| 296 |
+
"deployment",
|
| 297 |
+
"production",
|
| 298 |
+
"hosting",
|
| 299 |
+
"vercel",
|
| 300 |
+
"netlify",
|
| 301 |
+
"cloudflare",
|
| 302 |
+
"railway",
|
| 303 |
+
"render",
|
| 304 |
+
"heroku",
|
| 305 |
+
"fly.io",
|
| 306 |
+
"azure",
|
| 307 |
+
"aws",
|
| 308 |
+
"lambda",
|
| 309 |
+
"cloud run",
|
| 310 |
+
"gcp",
|
| 311 |
+
"kubernetes",
|
| 312 |
+
"k8s",
|
| 313 |
+
"sagemaker",
|
| 314 |
+
"app service",
|
| 315 |
+
"spaces",
|
| 316 |
+
"inference endpoint",
|
| 317 |
+
"ecs",
|
| 318 |
+
"supabase",
|
| 319 |
+
"serverless"
|
| 320 |
+
]
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"repo": "open-webui/open-webui",
|
| 324 |
+
"group": "framework",
|
| 325 |
+
"title_terms": [
|
| 326 |
+
"deploy",
|
| 327 |
+
"deployed",
|
| 328 |
+
"deployment",
|
| 329 |
+
"production",
|
| 330 |
+
"hosting",
|
| 331 |
+
"vercel",
|
| 332 |
+
"netlify",
|
| 333 |
+
"cloudflare",
|
| 334 |
+
"railway",
|
| 335 |
+
"render",
|
| 336 |
+
"heroku",
|
| 337 |
+
"fly.io",
|
| 338 |
+
"azure",
|
| 339 |
+
"aws",
|
| 340 |
+
"lambda",
|
| 341 |
+
"cloud run",
|
| 342 |
+
"gcp",
|
| 343 |
+
"kubernetes",
|
| 344 |
+
"k8s",
|
| 345 |
+
"sagemaker",
|
| 346 |
+
"app service",
|
| 347 |
+
"spaces",
|
| 348 |
+
"inference endpoint",
|
| 349 |
+
"ecs",
|
| 350 |
+
"supabase",
|
| 351 |
+
"serverless"
|
| 352 |
+
]
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"repo": "huggingface/text-generation-inference",
|
| 356 |
+
"group": "framework",
|
| 357 |
+
"title_terms": [
|
| 358 |
+
"deploy",
|
| 359 |
+
"deployed",
|
| 360 |
+
"deployment",
|
| 361 |
+
"production",
|
| 362 |
+
"hosting",
|
| 363 |
+
"vercel",
|
| 364 |
+
"netlify",
|
| 365 |
+
"cloudflare",
|
| 366 |
+
"railway",
|
| 367 |
+
"render",
|
| 368 |
+
"heroku",
|
| 369 |
+
"fly.io",
|
| 370 |
+
"azure",
|
| 371 |
+
"aws",
|
| 372 |
+
"lambda",
|
| 373 |
+
"cloud run",
|
| 374 |
+
"gcp",
|
| 375 |
+
"kubernetes",
|
| 376 |
+
"k8s",
|
| 377 |
+
"sagemaker",
|
| 378 |
+
"app service",
|
| 379 |
+
"spaces",
|
| 380 |
+
"inference endpoint",
|
| 381 |
+
"ecs",
|
| 382 |
+
"supabase",
|
| 383 |
+
"serverless"
|
| 384 |
+
]
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"repo": "langgenius/dify",
|
| 388 |
+
"group": "framework",
|
| 389 |
+
"title_terms": [
|
| 390 |
+
"deploy",
|
| 391 |
+
"deployed",
|
| 392 |
+
"deployment",
|
| 393 |
+
"production",
|
| 394 |
+
"hosting",
|
| 395 |
+
"vercel",
|
| 396 |
+
"netlify",
|
| 397 |
+
"cloudflare",
|
| 398 |
+
"railway",
|
| 399 |
+
"render",
|
| 400 |
+
"heroku",
|
| 401 |
+
"fly.io",
|
| 402 |
+
"azure",
|
| 403 |
+
"aws",
|
| 404 |
+
"lambda",
|
| 405 |
+
"cloud run",
|
| 406 |
+
"gcp",
|
| 407 |
+
"kubernetes",
|
| 408 |
+
"k8s",
|
| 409 |
+
"sagemaker",
|
| 410 |
+
"app service",
|
| 411 |
+
"spaces",
|
| 412 |
+
"inference endpoint",
|
| 413 |
+
"ecs",
|
| 414 |
+
"supabase",
|
| 415 |
+
"serverless"
|
| 416 |
+
]
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"repo": "infiniflow/ragflow",
|
| 420 |
+
"group": "framework",
|
| 421 |
+
"title_terms": [
|
| 422 |
+
"deploy",
|
| 423 |
+
"deployed",
|
| 424 |
+
"deployment",
|
| 425 |
+
"production",
|
| 426 |
+
"hosting",
|
| 427 |
+
"vercel",
|
| 428 |
+
"netlify",
|
| 429 |
+
"cloudflare",
|
| 430 |
+
"railway",
|
| 431 |
+
"render",
|
| 432 |
+
"heroku",
|
| 433 |
+
"fly.io",
|
| 434 |
+
"azure",
|
| 435 |
+
"aws",
|
| 436 |
+
"lambda",
|
| 437 |
+
"cloud run",
|
| 438 |
+
"gcp",
|
| 439 |
+
"kubernetes",
|
| 440 |
+
"k8s",
|
| 441 |
+
"sagemaker",
|
| 442 |
+
"app service",
|
| 443 |
+
"spaces",
|
| 444 |
+
"inference endpoint",
|
| 445 |
+
"ecs",
|
| 446 |
+
"supabase",
|
| 447 |
+
"serverless"
|
| 448 |
+
]
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"repo": "Mintplex-Labs/anything-llm",
|
| 452 |
+
"group": "framework",
|
| 453 |
+
"title_terms": [
|
| 454 |
+
"deploy",
|
| 455 |
+
"deployed",
|
| 456 |
+
"deployment",
|
| 457 |
+
"production",
|
| 458 |
+
"hosting",
|
| 459 |
+
"vercel",
|
| 460 |
+
"netlify",
|
| 461 |
+
"cloudflare",
|
| 462 |
+
"railway",
|
| 463 |
+
"render",
|
| 464 |
+
"heroku",
|
| 465 |
+
"fly.io",
|
| 466 |
+
"azure",
|
| 467 |
+
"aws",
|
| 468 |
+
"lambda",
|
| 469 |
+
"cloud run",
|
| 470 |
+
"gcp",
|
| 471 |
+
"kubernetes",
|
| 472 |
+
"k8s",
|
| 473 |
+
"sagemaker",
|
| 474 |
+
"app service",
|
| 475 |
+
"spaces",
|
| 476 |
+
"inference endpoint",
|
| 477 |
+
"ecs",
|
| 478 |
+
"supabase",
|
| 479 |
+
"serverless"
|
| 480 |
+
]
|
| 481 |
+
},
|
| 482 |
+
{
|
| 483 |
+
"repo": "huggingface/chat-ui",
|
| 484 |
+
"group": "framework",
|
| 485 |
+
"title_terms": [
|
| 486 |
+
"deploy",
|
| 487 |
+
"deployed",
|
| 488 |
+
"deployment",
|
| 489 |
+
"production",
|
| 490 |
+
"hosting",
|
| 491 |
+
"vercel",
|
| 492 |
+
"netlify",
|
| 493 |
+
"cloudflare",
|
| 494 |
+
"railway",
|
| 495 |
+
"render",
|
| 496 |
+
"heroku",
|
| 497 |
+
"fly.io",
|
| 498 |
+
"azure",
|
| 499 |
+
"aws",
|
| 500 |
+
"lambda",
|
| 501 |
+
"cloud run",
|
| 502 |
+
"gcp",
|
| 503 |
+
"kubernetes",
|
| 504 |
+
"k8s",
|
| 505 |
+
"sagemaker",
|
| 506 |
+
"app service",
|
| 507 |
+
"spaces",
|
| 508 |
+
"inference endpoint",
|
| 509 |
+
"ecs",
|
| 510 |
+
"supabase",
|
| 511 |
+
"serverless"
|
| 512 |
+
]
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"repo": "vercel/chatbot",
|
| 516 |
+
"group": "framework",
|
| 517 |
+
"title_terms": [
|
| 518 |
+
"deploy",
|
| 519 |
+
"deployed",
|
| 520 |
+
"deployment",
|
| 521 |
+
"production",
|
| 522 |
+
"hosting",
|
| 523 |
+
"vercel",
|
| 524 |
+
"netlify",
|
| 525 |
+
"cloudflare",
|
| 526 |
+
"railway",
|
| 527 |
+
"render",
|
| 528 |
+
"heroku",
|
| 529 |
+
"fly.io",
|
| 530 |
+
"azure",
|
| 531 |
+
"aws",
|
| 532 |
+
"lambda",
|
| 533 |
+
"cloud run",
|
| 534 |
+
"gcp",
|
| 535 |
+
"kubernetes",
|
| 536 |
+
"k8s",
|
| 537 |
+
"sagemaker",
|
| 538 |
+
"app service",
|
| 539 |
+
"spaces",
|
| 540 |
+
"inference endpoint",
|
| 541 |
+
"ecs",
|
| 542 |
+
"supabase",
|
| 543 |
+
"serverless"
|
| 544 |
+
]
|
| 545 |
+
},
|
| 546 |
+
{
|
| 547 |
+
"repo": "BerriAI/litellm",
|
| 548 |
+
"group": "framework",
|
| 549 |
+
"title_terms": [
|
| 550 |
+
"deploy",
|
| 551 |
+
"deployed",
|
| 552 |
+
"deployment",
|
| 553 |
+
"production",
|
| 554 |
+
"hosting",
|
| 555 |
+
"vercel",
|
| 556 |
+
"netlify",
|
| 557 |
+
"cloudflare",
|
| 558 |
+
"railway",
|
| 559 |
+
"render",
|
| 560 |
+
"heroku",
|
| 561 |
+
"fly.io",
|
| 562 |
+
"azure",
|
| 563 |
+
"aws",
|
| 564 |
+
"lambda",
|
| 565 |
+
"cloud run",
|
| 566 |
+
"gcp",
|
| 567 |
+
"kubernetes",
|
| 568 |
+
"k8s",
|
| 569 |
+
"sagemaker",
|
| 570 |
+
"app service",
|
| 571 |
+
"spaces",
|
| 572 |
+
"inference endpoint",
|
| 573 |
+
"ecs",
|
| 574 |
+
"supabase",
|
| 575 |
+
"serverless"
|
| 576 |
+
]
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"repo": "vllm-project/vllm",
|
| 580 |
+
"group": "framework",
|
| 581 |
+
"title_terms": [
|
| 582 |
+
"deploy",
|
| 583 |
+
"deployment",
|
| 584 |
+
"production",
|
| 585 |
+
"kubernetes",
|
| 586 |
+
"k8s",
|
| 587 |
+
"aws",
|
| 588 |
+
"sagemaker",
|
| 589 |
+
"gke",
|
| 590 |
+
"eks",
|
| 591 |
+
"cloud run",
|
| 592 |
+
"azure",
|
| 593 |
+
"helm",
|
| 594 |
+
"modal",
|
| 595 |
+
"runpod"
|
| 596 |
+
]
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"repo": "mastra-ai/mastra",
|
| 600 |
+
"group": "framework",
|
| 601 |
+
"title_terms": [
|
| 602 |
+
"deploy",
|
| 603 |
+
"deployed",
|
| 604 |
+
"deployment",
|
| 605 |
+
"production",
|
| 606 |
+
"hosting",
|
| 607 |
+
"vercel",
|
| 608 |
+
"netlify",
|
| 609 |
+
"cloudflare",
|
| 610 |
+
"railway",
|
| 611 |
+
"render",
|
| 612 |
+
"heroku",
|
| 613 |
+
"fly.io",
|
| 614 |
+
"azure",
|
| 615 |
+
"aws",
|
| 616 |
+
"lambda",
|
| 617 |
+
"cloud run",
|
| 618 |
+
"gcp",
|
| 619 |
+
"kubernetes",
|
| 620 |
+
"k8s",
|
| 621 |
+
"sagemaker",
|
| 622 |
+
"app service",
|
| 623 |
+
"spaces",
|
| 624 |
+
"inference endpoint",
|
| 625 |
+
"ecs",
|
| 626 |
+
"supabase",
|
| 627 |
+
"serverless"
|
| 628 |
+
]
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"repo": "google/adk-python",
|
| 632 |
+
"group": "framework",
|
| 633 |
+
"title_terms": [
|
| 634 |
+
"deploy",
|
| 635 |
+
"deployed",
|
| 636 |
+
"deployment",
|
| 637 |
+
"production",
|
| 638 |
+
"hosting",
|
| 639 |
+
"vercel",
|
| 640 |
+
"netlify",
|
| 641 |
+
"cloudflare",
|
| 642 |
+
"railway",
|
| 643 |
+
"render",
|
| 644 |
+
"heroku",
|
| 645 |
+
"fly.io",
|
| 646 |
+
"azure",
|
| 647 |
+
"aws",
|
| 648 |
+
"lambda",
|
| 649 |
+
"cloud run",
|
| 650 |
+
"gcp",
|
| 651 |
+
"kubernetes",
|
| 652 |
+
"k8s",
|
| 653 |
+
"sagemaker",
|
| 654 |
+
"app service",
|
| 655 |
+
"spaces",
|
| 656 |
+
"inference endpoint",
|
| 657 |
+
"ecs",
|
| 658 |
+
"supabase",
|
| 659 |
+
"serverless"
|
| 660 |
+
]
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"repo": "crewAIInc/crewAI",
|
| 664 |
+
"group": "framework",
|
| 665 |
+
"title_terms": [
|
| 666 |
+
"deploy",
|
| 667 |
+
"deployed",
|
| 668 |
+
"deployment",
|
| 669 |
+
"production",
|
| 670 |
+
"hosting",
|
| 671 |
+
"vercel",
|
| 672 |
+
"netlify",
|
| 673 |
+
"cloudflare",
|
| 674 |
+
"railway",
|
| 675 |
+
"render",
|
| 676 |
+
"heroku",
|
| 677 |
+
"fly.io",
|
| 678 |
+
"azure",
|
| 679 |
+
"aws",
|
| 680 |
+
"lambda",
|
| 681 |
+
"cloud run",
|
| 682 |
+
"gcp",
|
| 683 |
+
"kubernetes",
|
| 684 |
+
"k8s",
|
| 685 |
+
"sagemaker",
|
| 686 |
+
"app service",
|
| 687 |
+
"spaces",
|
| 688 |
+
"inference endpoint",
|
| 689 |
+
"ecs",
|
| 690 |
+
"supabase",
|
| 691 |
+
"serverless"
|
| 692 |
+
]
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"repo": "agno-agi/agno",
|
| 696 |
+
"group": "framework",
|
| 697 |
+
"title_terms": [
|
| 698 |
+
"deploy",
|
| 699 |
+
"deployed",
|
| 700 |
+
"deployment",
|
| 701 |
+
"production",
|
| 702 |
+
"hosting",
|
| 703 |
+
"vercel",
|
| 704 |
+
"netlify",
|
| 705 |
+
"cloudflare",
|
| 706 |
+
"railway",
|
| 707 |
+
"render",
|
| 708 |
+
"heroku",
|
| 709 |
+
"fly.io",
|
| 710 |
+
"azure",
|
| 711 |
+
"aws",
|
| 712 |
+
"lambda",
|
| 713 |
+
"cloud run",
|
| 714 |
+
"gcp",
|
| 715 |
+
"kubernetes",
|
| 716 |
+
"k8s",
|
| 717 |
+
"sagemaker",
|
| 718 |
+
"app service",
|
| 719 |
+
"spaces",
|
| 720 |
+
"inference endpoint",
|
| 721 |
+
"ecs",
|
| 722 |
+
"supabase",
|
| 723 |
+
"serverless"
|
| 724 |
+
]
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"repo": "Azure-Samples/azure-search-openai-demo",
|
| 728 |
+
"group": "sample",
|
| 729 |
+
"title_terms": [
|
| 730 |
+
"deploy",
|
| 731 |
+
"deployed",
|
| 732 |
+
"deployment",
|
| 733 |
+
"production",
|
| 734 |
+
"hosting",
|
| 735 |
+
"vercel",
|
| 736 |
+
"netlify",
|
| 737 |
+
"cloudflare",
|
| 738 |
+
"railway",
|
| 739 |
+
"render",
|
| 740 |
+
"heroku",
|
| 741 |
+
"fly.io",
|
| 742 |
+
"azure",
|
| 743 |
+
"aws",
|
| 744 |
+
"lambda",
|
| 745 |
+
"cloud run",
|
| 746 |
+
"gcp",
|
| 747 |
+
"kubernetes",
|
| 748 |
+
"k8s",
|
| 749 |
+
"sagemaker",
|
| 750 |
+
"app service",
|
| 751 |
+
"spaces",
|
| 752 |
+
"inference endpoint",
|
| 753 |
+
"ecs",
|
| 754 |
+
"supabase",
|
| 755 |
+
"serverless"
|
| 756 |
+
]
|
| 757 |
+
},
|
| 758 |
+
{
|
| 759 |
+
"repo": "Azure-Samples/chat-with-your-data-solution-accelerator",
|
| 760 |
+
"group": "sample",
|
| 761 |
+
"title_terms": [
|
| 762 |
+
"deploy",
|
| 763 |
+
"deployed",
|
| 764 |
+
"deployment",
|
| 765 |
+
"production",
|
| 766 |
+
"hosting",
|
| 767 |
+
"vercel",
|
| 768 |
+
"netlify",
|
| 769 |
+
"cloudflare",
|
| 770 |
+
"railway",
|
| 771 |
+
"render",
|
| 772 |
+
"heroku",
|
| 773 |
+
"fly.io",
|
| 774 |
+
"azure",
|
| 775 |
+
"aws",
|
| 776 |
+
"lambda",
|
| 777 |
+
"cloud run",
|
| 778 |
+
"gcp",
|
| 779 |
+
"kubernetes",
|
| 780 |
+
"k8s",
|
| 781 |
+
"sagemaker",
|
| 782 |
+
"app service",
|
| 783 |
+
"spaces",
|
| 784 |
+
"inference endpoint",
|
| 785 |
+
"ecs",
|
| 786 |
+
"supabase",
|
| 787 |
+
"serverless"
|
| 788 |
+
]
|
| 789 |
+
},
|
| 790 |
+
{
|
| 791 |
+
"repo": "microsoft/sample-app-aoai-chatGPT",
|
| 792 |
+
"group": "sample",
|
| 793 |
+
"title_terms": [
|
| 794 |
+
"deploy",
|
| 795 |
+
"deployed",
|
| 796 |
+
"deployment",
|
| 797 |
+
"production",
|
| 798 |
+
"hosting",
|
| 799 |
+
"vercel",
|
| 800 |
+
"netlify",
|
| 801 |
+
"cloudflare",
|
| 802 |
+
"railway",
|
| 803 |
+
"render",
|
| 804 |
+
"heroku",
|
| 805 |
+
"fly.io",
|
| 806 |
+
"azure",
|
| 807 |
+
"aws",
|
| 808 |
+
"lambda",
|
| 809 |
+
"cloud run",
|
| 810 |
+
"gcp",
|
| 811 |
+
"kubernetes",
|
| 812 |
+
"k8s",
|
| 813 |
+
"sagemaker",
|
| 814 |
+
"app service",
|
| 815 |
+
"spaces",
|
| 816 |
+
"inference endpoint",
|
| 817 |
+
"ecs",
|
| 818 |
+
"supabase",
|
| 819 |
+
"serverless"
|
| 820 |
+
]
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"repo": "GoogleCloudPlatform/generative-ai",
|
| 824 |
+
"group": "sample",
|
| 825 |
+
"title_terms": [
|
| 826 |
+
"deploy",
|
| 827 |
+
"deployed",
|
| 828 |
+
"deployment",
|
| 829 |
+
"production",
|
| 830 |
+
"hosting",
|
| 831 |
+
"vercel",
|
| 832 |
+
"netlify",
|
| 833 |
+
"cloudflare",
|
| 834 |
+
"railway",
|
| 835 |
+
"render",
|
| 836 |
+
"heroku",
|
| 837 |
+
"fly.io",
|
| 838 |
+
"azure",
|
| 839 |
+
"aws",
|
| 840 |
+
"lambda",
|
| 841 |
+
"cloud run",
|
| 842 |
+
"gcp",
|
| 843 |
+
"kubernetes",
|
| 844 |
+
"k8s",
|
| 845 |
+
"sagemaker",
|
| 846 |
+
"app service",
|
| 847 |
+
"spaces",
|
| 848 |
+
"inference endpoint",
|
| 849 |
+
"ecs",
|
| 850 |
+
"supabase",
|
| 851 |
+
"serverless"
|
| 852 |
+
]
|
| 853 |
+
},
|
| 854 |
+
{
|
| 855 |
+
"repo": "GoogleCloudPlatform/agent-starter-pack",
|
| 856 |
+
"group": "sample",
|
| 857 |
+
"title_terms": [
|
| 858 |
+
"deploy",
|
| 859 |
+
"deployed",
|
| 860 |
+
"deployment",
|
| 861 |
+
"production",
|
| 862 |
+
"hosting",
|
| 863 |
+
"vercel",
|
| 864 |
+
"netlify",
|
| 865 |
+
"cloudflare",
|
| 866 |
+
"railway",
|
| 867 |
+
"render",
|
| 868 |
+
"heroku",
|
| 869 |
+
"fly.io",
|
| 870 |
+
"azure",
|
| 871 |
+
"aws",
|
| 872 |
+
"lambda",
|
| 873 |
+
"cloud run",
|
| 874 |
+
"gcp",
|
| 875 |
+
"kubernetes",
|
| 876 |
+
"k8s",
|
| 877 |
+
"sagemaker",
|
| 878 |
+
"app service",
|
| 879 |
+
"spaces",
|
| 880 |
+
"inference endpoint",
|
| 881 |
+
"ecs",
|
| 882 |
+
"supabase",
|
| 883 |
+
"serverless"
|
| 884 |
+
]
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"repo": "Azure-Samples/contoso-chat",
|
| 888 |
+
"group": "sample",
|
| 889 |
+
"title_terms": [
|
| 890 |
+
"deploy",
|
| 891 |
+
"deployed",
|
| 892 |
+
"deployment",
|
| 893 |
+
"production",
|
| 894 |
+
"hosting",
|
| 895 |
+
"vercel",
|
| 896 |
+
"netlify",
|
| 897 |
+
"cloudflare",
|
| 898 |
+
"railway",
|
| 899 |
+
"render",
|
| 900 |
+
"heroku",
|
| 901 |
+
"fly.io",
|
| 902 |
+
"azure",
|
| 903 |
+
"aws",
|
| 904 |
+
"lambda",
|
| 905 |
+
"cloud run",
|
| 906 |
+
"gcp",
|
| 907 |
+
"kubernetes",
|
| 908 |
+
"k8s",
|
| 909 |
+
"sagemaker",
|
| 910 |
+
"app service",
|
| 911 |
+
"spaces",
|
| 912 |
+
"inference endpoint",
|
| 913 |
+
"ecs",
|
| 914 |
+
"supabase",
|
| 915 |
+
"serverless"
|
| 916 |
+
]
|
| 917 |
+
},
|
| 918 |
+
{
|
| 919 |
+
"repo": "Azure-Samples/rag-postgres-openai-python",
|
| 920 |
+
"group": "sample",
|
| 921 |
+
"title_terms": [
|
| 922 |
+
"deploy",
|
| 923 |
+
"deployed",
|
| 924 |
+
"deployment",
|
| 925 |
+
"production",
|
| 926 |
+
"hosting",
|
| 927 |
+
"vercel",
|
| 928 |
+
"netlify",
|
| 929 |
+
"cloudflare",
|
| 930 |
+
"railway",
|
| 931 |
+
"render",
|
| 932 |
+
"heroku",
|
| 933 |
+
"fly.io",
|
| 934 |
+
"azure",
|
| 935 |
+
"aws",
|
| 936 |
+
"lambda",
|
| 937 |
+
"cloud run",
|
| 938 |
+
"gcp",
|
| 939 |
+
"kubernetes",
|
| 940 |
+
"k8s",
|
| 941 |
+
"sagemaker",
|
| 942 |
+
"app service",
|
| 943 |
+
"spaces",
|
| 944 |
+
"inference endpoint",
|
| 945 |
+
"ecs",
|
| 946 |
+
"supabase",
|
| 947 |
+
"serverless"
|
| 948 |
+
]
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"repo": "stackblitz-labs/bolt.diy",
|
| 952 |
+
"group": "builder",
|
| 953 |
+
"title_terms": [
|
| 954 |
+
"deploy",
|
| 955 |
+
"netlify",
|
| 956 |
+
"vercel",
|
| 957 |
+
"cloudflare",
|
| 958 |
+
"deployment",
|
| 959 |
+
"publish",
|
| 960 |
+
"hosting",
|
| 961 |
+
"coolify",
|
| 962 |
+
"production"
|
| 963 |
+
]
|
| 964 |
+
},
|
| 965 |
+
{
|
| 966 |
+
"repo": "dyad-sh/dyad",
|
| 967 |
+
"group": "builder",
|
| 968 |
+
"title_terms": [
|
| 969 |
+
"deploy",
|
| 970 |
+
"vercel",
|
| 971 |
+
"netlify",
|
| 972 |
+
"supabase",
|
| 973 |
+
"publish",
|
| 974 |
+
"deployment",
|
| 975 |
+
"production"
|
| 976 |
+
]
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"repo": "anthropics/claude-code",
|
| 980 |
+
"group": "builder",
|
| 981 |
+
"title_terms": [
|
| 982 |
+
"deploy",
|
| 983 |
+
"vercel",
|
| 984 |
+
"netlify",
|
| 985 |
+
"railway",
|
| 986 |
+
"cloudflare",
|
| 987 |
+
"fly.io",
|
| 988 |
+
"render",
|
| 989 |
+
"supabase",
|
| 990 |
+
"heroku",
|
| 991 |
+
"production deploy"
|
| 992 |
+
]
|
| 993 |
+
}
|
| 994 |
+
]
|
| 995 |
+
}
|
scripts/sample_for_review.py
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Random sample of coded threads for manual review (local only; prints titles/excerpts).
|
| 3 |
+
Usage: python3 -I scripts/sample_for_review.py coded_kw.jsonl N [seed] [--exclude judgement.csv ...]
|
| 4 |
+
Prints one block per thread plus a CSV header to fill in manual/judgement_full.csv (keyed by url)."""
|
| 5 |
+
import json, random, sys, csv
|
| 6 |
+
args = sys.argv[1:]
|
| 7 |
+
excl = set()
|
| 8 |
+
if "--exclude" in args:
|
| 9 |
+
i = args.index("--exclude")
|
| 10 |
+
for p in args[i + 1:]:
|
| 11 |
+
excl |= {r.get("url") for r in csv.DictReader(open(p)) if r.get("url")}
|
| 12 |
+
args = args[:i]
|
| 13 |
+
coded = [json.loads(l) for l in open(args[0])]
|
| 14 |
+
n, seed = int(args[1]), int(args[2]) if len(args) > 2 else 20261028
|
| 15 |
+
pool = [c for c in coded if c["url"] not in excl]
|
| 16 |
+
random.Random(seed).shuffle(pool)
|
| 17 |
+
for c in pool[:n]:
|
| 18 |
+
print(f'{c["url"]} [{c["state"]}/{c["state_reason"]}] T: {c["title"]}')
|
| 19 |
+
print(f' E: {c["excerpt_private"]}')
|
| 20 |
+
print(f' KW rel={c["kw_relevant"]} host={c["kw_target_host"]} app={c["kw_app_type"]} fw={c["kw_framework"]} '
|
| 21 |
+
f'prov={c["kw_model_provider"]} cat={c["kw_topic"]} err={c["kw_error_type"]} '
|
| 22 |
+
f'res={c["kw_resolution_type"]} aigen={c["kw_ai_generated_app"]}/{c["kw_ai_tool"]}')
|
scripts/stats.py
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""Descriptive numbers for a threads CSV (published format).
|
| 3 |
+
Usage: python3 -I scripts/stats.py data/threads.csv [--relevant-only]"""
|
| 4 |
+
import csv, sys, statistics, collections
|
| 5 |
+
|
| 6 |
+
rows = list(csv.DictReader(open(sys.argv[1])))
|
| 7 |
+
if "--relevant-only" in sys.argv and rows and "relevance" in rows[0]:
|
| 8 |
+
rows = [r for r in rows if r["relevance"] == "yes"]
|
| 9 |
+
n = len(rows)
|
| 10 |
+
print(f"threads: {n}")
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
def pct(k, tot):
|
| 14 |
+
return f"{100 * k / tot:.0f}%" if tot else "-"
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
def med(vals):
|
| 18 |
+
v = [float(x) for x in vals if x not in ("", None)]
|
| 19 |
+
return (f"{statistics.median(v):.1f} h (n={len(v)})") if v else "-"
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
print(f"resolved: {pct(sum(r['resolved'] == 'yes' for r in rows), n)}")
|
| 23 |
+
print(f"maintainer replied: {pct(sum(r['maintainer_replied'] == 'yes' for r in rows), n)}")
|
| 24 |
+
print(f"no reply at all (no time_to_first_response): {pct(sum(r['time_to_first_response_h'] == '' for r in rows), n)}")
|
| 25 |
+
print(f"median time to first response: {med(r['time_to_first_response_h'] for r in rows)}")
|
| 26 |
+
print(f"median time to close (closed only): {med(r['time_to_close_h'] for r in rows)}")
|
| 27 |
+
for col in ["thread_type", "repo_group", "target_host", "app_type", "framework", "topic", "error_type",
|
| 28 |
+
"resolution_type", "ai_generated_app", "ai_tool", "coding_confidence"]:
|
| 29 |
+
c = collections.Counter(r[col] or "(empty)" for r in rows)
|
| 30 |
+
print(f"\n{col}: " + ", ".join(f"{k} {v} ({pct(v, n)})" for k, v in c.most_common(20)))
|
| 31 |
+
prov = collections.Counter(p for r in rows for p in (r["model_provider"] or "unspecified").split(";"))
|
| 32 |
+
print("\nmodel_provider (multi): " + ", ".join(f"{k} {v}" for k, v in prov.most_common(12)))
|
| 33 |
+
# topic x resolution / response, for the categories with >= 20 threads
|
| 34 |
+
print("\nby topic: n, resolved, median first response")
|
| 35 |
+
by = collections.defaultdict(list)
|
| 36 |
+
for r in rows:
|
| 37 |
+
by[r["topic"]].append(r)
|
| 38 |
+
for k, lst in sorted(by.items(), key=lambda kv: -len(kv[1])):
|
| 39 |
+
if len(lst) >= 20:
|
| 40 |
+
print(f" {k:22s} {len(lst):5d} {pct(sum(x['resolved'] == 'yes' for x in lst), len(lst)):>5s} {med(x['time_to_first_response_h'] for x in lst)}")
|
| 41 |
+
print("\nby target_host (>= 20): n, resolved, maintainer replied, median first response")
|
| 42 |
+
by = collections.defaultdict(list)
|
| 43 |
+
for r in rows:
|
| 44 |
+
by[r["target_host"]].append(r)
|
| 45 |
+
for k, lst in sorted(by.items(), key=lambda kv: -len(kv[1])):
|
| 46 |
+
if len(lst) >= 20:
|
| 47 |
+
print(f" {k:22s} {len(lst):5d} {pct(sum(x['resolved'] == 'yes' for x in lst), len(lst)):>5s} "
|
| 48 |
+
f"{pct(sum(x['maintainer_replied'] == 'yes' for x in lst), len(lst)):>5s} {med(x['time_to_first_response_h'] for x in lst)}")
|
| 49 |
+
if rows and "thread_type" in rows[0]:
|
| 50 |
+
types = ["problem", "feature", "docs", "question", "pr_ci", "other"]
|
| 51 |
+
print("\nby thread_type: n, resolved, maintainer replied, no reply, median first response")
|
| 52 |
+
for tt in types:
|
| 53 |
+
lst = [r for r in rows if r["thread_type"] == tt]
|
| 54 |
+
if lst:
|
| 55 |
+
print(f" {tt:10s} {len(lst):5d} {pct(sum(x['resolved'] == 'yes' for x in lst), len(lst)):>5s} "
|
| 56 |
+
f"{pct(sum(x['maintainer_replied'] == 'yes' for x in lst), len(lst)):>5s} "
|
| 57 |
+
f"{pct(sum(x['time_to_first_response_h'] == '' for x in lst), len(lst)):>5s} {med(x['time_to_first_response_h'] for x in lst)}")
|
| 58 |
+
print("\nthread_type by target_host (hosts >= 20): n, then share of each type")
|
| 59 |
+
by = collections.defaultdict(list)
|
| 60 |
+
for r in rows:
|
| 61 |
+
by[r["target_host"]].append(r)
|
| 62 |
+
print(" " + " " * 22 + " n " + " ".join(f"{t:>8s}" for t in types))
|
| 63 |
+
for k, lst in sorted(by.items(), key=lambda kv: -len(kv[1])):
|
| 64 |
+
if len(lst) >= 20:
|
| 65 |
+
print(f" {k:22s} {len(lst):5d} " + " ".join(f"{pct(sum(x['thread_type'] == t for x in lst), len(lst)):>8s}" for t in types))
|
| 66 |
+
for tt in ("feature", "docs"):
|
| 67 |
+
c = collections.Counter(r["target_host"] for r in rows if r["thread_type"] == tt)
|
| 68 |
+
print(f"\nhosts with the most {tt} threads: " + ", ".join(f"{k} {v}" for k, v in c.most_common(8)))
|
| 69 |
+
print("\nthread_type by repo_group: " + "; ".join(
|
| 70 |
+
f"{g}: " + ", ".join(f"{t} {pct(sum(x['thread_type'] == t for x in rows if x['repo_group'] == g), sum(x['repo_group'] == g for x in rows))}" for t in types)
|
| 71 |
+
for g in sorted({r['repo_group'] for r in rows})))
|
| 72 |
+
print("\nby year: n")
|
| 73 |
+
c = collections.Counter(r["created_at"][:4] for r in rows)
|
| 74 |
+
print(" " + ", ".join(f"{k}: {v}" for k, v in sorted(c.items())))
|