Introduce SparkBotLLM and its local AI setup workflow

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 SparkBotLLM-1.7B-Q4_0.gguf filter=lfs diff=lfs merge=lfs -text
+assets/sparkbot.webp filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,49 +1,90 @@
 ---
 library_name: peft
 base_model: LiquidAI/LFM3-1.7B
-tags: [dgx-spark, sparkbot, gguf, lora, text-generation]
+tags: [dgx-spark, sparkbot, gguf, lora, text-generation, local-ai]
 license: other
 license_name: lfm-open-license-v1
 license_link: LICENSE
 language: [en]
 pipeline_tag: text-generation
 ---
-# SparkBotLLM 1.7B
+<div align="center">
+<img src="assets/sparkbot.webp" alt="Sparkbot" width="220">
 
-A local assistant for DGX Spark setup and local inference, adapted from LiquidAI LFM3 1.7B. Sparkbot supplies the live inspection tools, reviewed actions and dated support references. Model inference can run privately on CPU with a 4K context and one request at a time.
+# SparkBotLLM
 
-SparkBotLLM helps you take the next step with your Spark: check a setup, understand a deployment requirement, or inspect a problem before changing anything. The surrounding Sparkbot dashboard gives it hardware inspection and terminal tools, with owner review for changes.
+### Your first local assistant. Built to help you run the next one.
 
-For current releases, use discovery tools and primary sources. For live clocks, memory and Docker status, use hardware tools. The weights alone cannot inspect your machine.
+[**Discover Sparkbot**](https://sparkbot.dev) · [**Download Q4**](https://huggingface.co/oxbyte8/SparkBotLLM-1.7B/tree/main) · [**LiquidAI base model**](https://huggingface.co/LiquidAI/LFM3-1.7B)
+</div>
 
-## Getting started
+Getting a powerful model running locally should feel approachable. SparkBotLLM gives DGX Spark owners a small assistant to help them get there: understand their hardware, work through setup, choose a deployment profile and troubleshoot the next step.
 
-The model is designed for the [Sparkbot dashboard](https://github.com/x00byte/DGX-Sparkbot). The dashboard and website installation package are being prepared for release; the source repository is currently private. You can run the GGUF independently with a current LFM3-capable llama.cpp build using the command below. No Docker or GPU is required for CPU inference.
+It is the local assistant inside **[Sparkbot](https://sparkbot.dev)**, a dashboard for managing NVIDIA DGX Sparks. The idea is simple: start with a compact AI helper on your own machine, then use it to help deploy and manage larger local models. Beginners get clear guidance. Experienced users keep access to their terminal and configuration.
 
-Download the GGUF from **Files and versions**. It is approximately 1.05 GB. Start with a 4K context, four CPU threads and one request at a time. This keeps the helper small while your main model uses the GPU.
+## Small enough to stay alongside your main model
 
-## Files and use
+| | SparkBotLLM |
+|---|---|
+| Base | LiquidAI LFM3 1.7B |
+| Download | Approximately 1.05 GB, Q4_0 GGUF |
+| Default execution | Local CPU, four threads, one request at a time |
+| Serving context | 4,096 tokens |
+| Sparkbot runtime | 9.2 MB native CPU package; no Docker or PyTorch required |
+| Helper endpoint | `127.0.0.1:10004` |
+| Dashboard | Port `10005` |
 
-`SparkBotLLM-1.7B-Q4_0.gguf` is the merged, CPU-ready Q4_0 model with Q6_K token embeddings. The adapter files require the exact BF16 base revision `ac2586a34f1a1c03329dbdcdefabd8b6478ddaa7`; do not apply them to LiquidAI’s separate QAD quantized model. The GGUF embeds the original chat template. Disable thinking for short helper turns.
+CPU inference leaves GPU execution available for your main model. CPU work and unified-memory traffic still share the Spark’s resources. The dashboard caps its helper service at 3 GiB; that is a configured limit, not a claim of measured resident memory.
 
-Run with a recent LFM3-capable llama.cpp server, for example:
+## From a question to a useful next step
+
+Inside Sparkbot, the assistant can use tools to inspect Docker support, GPU clocks, memory and connections; discover new models; explain deployment requirements; and propose terminal commands. Changes wait for owner approval. Administrator passwords are entered in the dashboard’s approval flow and are not sent to the model.
+
+A typical journey is to check the Spark, resolve a missing dependency, select a bundled recipe, then monitor the running model and compare decode or prefill results. SparkBotLLM stays beside that workflow as local help.
+
+**The dashboard supplies these tools and controls.** Running the model file alone gives you a chat model; it does not grant shell access, fetch current releases or deploy models by itself. Model inference runs locally. Downloads and online discovery contact their respective sources.
+
+## Run it today
+
+Download `SparkBotLLM-1.7B-Q4_0.gguf` from **Files and versions**. It is public and needs no access request. Use a current LFM3-capable llama.cpp build:
 
 ```sh
-llama-server --model SparkBotLLM-1.7B-Q4_0.gguf --alias SparkBotLLM --host 127.0.0.1 --port 10004 --ctx-size 4096 --parallel 1 --threads 4 --n-gpu-layers 0 --jinja --chat-template-kwargs '{"enable_thinking":false}'
+llama-server \
+  --model SparkBotLLM-1.7B-Q4_0.gguf \
+  --alias SparkBotLLM \
+  --host 127.0.0.1 --port 10004 \
+  --ctx-size 4096 --parallel 1 --threads 4 \
+  --n-gpu-layers 0 --jinja \
+  --chat-template-kwargs '{"enable_thinking":false}'
 ```
 
-## What changed
+The GGUF includes the chat template. Start with a 4K context and short, tool-assisted turns. The Sparkbot website package also includes a compact runtime and guided setup. **sparkbot.dev is the planned home; its hosting and domain connection are being prepared.** The model can already be downloaded and run independently.
 
-The base model received a short supervised adaptation focused on DGX Spark support, local inference concepts and choosing the right inspection tool. The application also supplies a dated reference pack; that reference material can be updated independently of the model.
+## What we adapted
 
-SparkBotLLM is intended to explain requirements, diagnose issues and propose useful next steps. Shell access and deployment automation come from the host application, not from the model file itself.
+SparkBotLLM received a short supervised LoRA adaptation focused on DGX Spark support, local inference and choosing the appropriate inspection tool. The training set contains **242 authored synthetic support examples**, informed by **44 dated support references**. It contains no private user chats or credentials.
 
-## Adaptation and checks
+Sparkbot also supplies relevant support references at runtime. This lets documentation evolve separately from the weights. Live hardware facts come from inspection tools; new software and model releases should come from current primary sources.
 
-242 authored synthetic support examples, 44 dated support references, LoRA rank 8/alpha 16, assistant-only loss, AdamW learning rate 0.00005. A 192-step run selected step 128 by lowest validation loss. Six domain validation examples were used for selection: mean loss fell from 4.62 to 3.32. They are not an independent test set. Five independent Q4 tool-routing smoke checks passed after export. These narrow checks do not establish broad factual accuracy, safety or reliability. Peak training CUDA allocation was 3.59 GiB. Reports are included.
+| Training detail | Value |
+|---|---|
+| Exact base revision | `ac2586a34f1a1c03329dbdcdefabd8b6478ddaa7` |
+| LoRA | Rank 8, alpha 16 |
+| Optimizer | AdamW, learning rate 0.00005 |
+| Loss | Assistant tokens only |
+| Selected checkpoint | Step 128 of a 192-step run |
+| Peak CUDA allocation during training | 3.59 GiB |
 
-The adaptation covers setup, troubleshooting, tool use and local inference concepts. Training examples do not contain the owner’s private chats or credentials. Knowledge can be stale; web retrieval is supplied by the application, not the weights. The 4K serving limit is intentional; long-context quality was not evaluated.
+The adapter was merged into that exact BF16 base before Q4_0 conversion, with Q6_K token embeddings. **Do not apply this adapter to LiquidAI’s separate QAD quantized checkpoint.**
+
+## Checks and practical limits
+
+Six domain validation examples were used for checkpoint selection. Mean loss fell from **4.62 to 3.32**. Because they guided selection, these are validation results, not an independent final test. After Q4 export, **five independent tool-routing smoke checks passed**, covering hardware inspection, clocks, setup, recipes and model discovery. Reports are included in this repository.
+
+These are narrow checks. They do not establish broad factual accuracy or reliable deployment automation. SparkBotLLM is a small model and can make mistakes. Keep command review enabled, verify important guidance, and use tools for live measurements. Long-context quality beyond the 4K serving profile has not been evaluated.
 
 ## License and credits
 
-Based on LiquidAI LFM3 1.7B. The retained LFM Open License v1.0 applies, including its commercial conditions. SparkBotLLM is an adaptation, not a model trained from scratch. LiquidAI supplies the base weights and architecture; PyTorch, Transformers and PEFT support training; llama.cpp supplies GGUF conversion and the CPU runtime. Reference summaries credit NVIDIA, Docker, TensorFold and model publishers in the Sparkbot dataset. Sparkbot is an independent project by x00byte.
+SparkBotLLM is an adaptation of **[LiquidAI LFM3 1.7B](https://huggingface.co/LiquidAI/LFM3-1.7B)**, not a model trained from scratch. The retained **[LFM Open License v1.0](LICENSE)** applies, including its commercial conditions.
+
+LiquidAI supplies the base weights and architecture. PyTorch, Transformers and PEFT support training; llama.cpp supplies GGUF conversion and the compact CPU backend. Support references credit NVIDIA, Docker, TensorFold and the relevant model publishers. Sparkbot integration and supplied branding are by **x00byte**. Sparkbot is an independent project for DGX Spark owners, with no NVIDIA endorsement claimed.