## Problem
The Codex App cannot discover FCC provider/model slugs because it loads
a local model catalog at startup instead of querying FCC's `/v1/models`
endpoint. Users can configure FCC as a Responses provider, but its
native model picker remains incomplete. Fixes #1260.
## Changes
| Before | After |
| --- | --- |
| FCC generated the Codex catalog only when `fcc-codex` launched. | FCC
synchronizes the catalog from its canonical model lifecycle for
independently launched Codex clients. |
| Startup could leave a configured Codex App without a catalog file. |
Startup creates a missing catalog after routed-provider warming and
preserves an existing last-known-good catalog. |
| Every catalog publication replaced the file. | Identical content is
skipped and changed content is replaced atomically. |
| Catalog filesystem failures could escape a direct writer call. |
Runtime publication is fail-open and logs only the exception type. |
| Codex App setup was undocumented. | The README docum…
Add ChatGPT OAuth as an OpenAI Codex provider (#1279)
## Problem
FCC supports API-key and local providers, but customers cannot use
models included with their ChatGPT subscription. Reusing Codex
credentials would also couple FCC to another application's mutable
authentication state.
## Changes
| Before | After |
| --- | --- |
| OpenAI subscription access was unavailable. | Customers connect
ChatGPT from the local Admin UI with browser PKCE or device code. |
| Provider availability came only from static settings. | A
process-owned connected account adds, refreshes, and evicts OpenAI
models dynamically. |
| FCC had no ChatGPT Codex transport. | `openai/<model-id>` uses
FCC-owned renewable credentials and neutral Responses conversion. |
| Retries and reasoning continuity had no subscription-provider
coverage. | Commit-boundary retries prevent duplicate output while
encrypted reasoning and tool IDs survive round trips. |
<!-- greptile_comment -->
<details open><summary><h3>Greptile Summary</h3></summary>
This PR adds ChatGPT OAuth and a…
Make provider model discovery the sole catalog owner (#1221)
## Problem
Startup model-list I/O was split between a non-enforcing
configured-model validator and the real discovery path. Both queried
providers and populated the same cache even though synchronous cache
warm-up was the validator's only required effect.
## Changes
| Before | After |
| --- | --- |
| Validation and discovery independently resolved providers, queried
model lists, and cached results. | `ProviderModelDiscovery` solely owns
model-list queries, failure reporting, and cache population. |
| Startup ran configured-model validation and then launched discovery. |
Startup synchronously warms referenced providers through discovery, then
launches the existing missing-provider background pass. |
| Absence from a provider catalog produced a non-enforcing missing-model
warning. | Provider catalogs remain discovery metadata and provider
execution remains authoritative. |
| Configured model references retained environment-source metadata for
validator diagnostics. | Configured model …
Coordinate provider recovery under concurrent load (#1205)
## Problem
Concurrent transient failures could start independent retry, replay,
continuation, and repair loops while holding provider concurrency slots.
This multiplied upstream attempts and could delay or strand terminal
errors under fan-out.
## Changes
| Before | After |
| --- | --- |
| Retry paths owned separate attempt budgets. | One logical-execution
session caps all upstream work at five attempts. |
| Concurrent failures backed off independently. | One provider-owned
recovery episode elects a single half-open probe while followers
coalesce. |
| Backoff occupied stream concurrency. | Concurrency is held only while
an upstream operation or stream is active. |
| Provider catalog calls and stream creation used separate admission
paths. | Every upstream operation uses one provider-generation admission
controller. |
| Cancellation could leave recovery ownership or follower state
unresolved. | Cancellation releases permits, transfers probe ownership,
and unregisters waiting followers…
Add searchable model selection to the Admin UI (#1121)
## Problem
Admin model routing fields required users to construct provider-prefixed
model slugs. Optional tier overrides represented inheritance as an
unexplained blank value.
## Changes
| Before | After |
| --- | --- |
| Model inputs only gained suggestions after an individual provider
refresh. | Model inputs load configured and discovered canonical slugs
from one Admin catalog. |
| Model routing looked like unrestricted text entry. | Model routing
uses the browser's searchable model dropdown while retaining manual
entry. |
| Tier overrides displayed an empty value for fallback routing. | Tier
overrides display **None** and persist it as an unset override. |
| Model refresh returned provider-shaped cache internals. | Model
refresh returns the same canonical catalog consumed by the Admin UI. |
| Users inferred the provider/model slug format from examples. | The
Admin UI and README define and present complete provider/model slugs. |
<!-- greptile_comment -->
<details open><summary>…
Fix cross-drive messaging startup with native Claude plan storage (#1085)
## Problem
Messaging startup computed a relative path from `ALLOWED_DIR` to FCC's
plan directory. Windows raises before Telegram or Discord polling starts
when those paths are on different drives, while Claude requires custom
plan directories to remain inside the project root. Fixes #1077.
## Changes
| Before | After |
| --- | --- |
| Runtime startup converted FCC's plan path relative to the customer
workspace. | Runtime startup no longer computes or transports a custom
plan path. |
| Managed Claude configuration carried `plans_directory` through four
ownership layers. | Managed Claude relies on its native user-level plan
storage. |
| `~/.fcc/agent_workspace` was described as a Claude workspace. | The
existing path is explicitly owned as messaging state for
`sessions.json`. |
| Cross-volume `ALLOWED_DIR` values could disable messaging startup. |
`ALLOWED_DIR` remains the project workspace regardless of filesystem
volume. |
Disable nonessential Anthropic traffic for FCC Claude sessions (#1083)
## Problem
Claude proxy environment policy was duplicated between `fcc-claude`,
managed messaging, and live smoke drivers. Managed messaging preserved a
legacy endpoint variable, inherited more Anthropic state, and did not
disable nonessential traffic, so FCC-launched Claude sessions could
drift apart.
## Changes
| Before | After |
| --- | --- |
| `fcc-claude` and managed messaging assembled proxy environments
independently. | One shared owner strips inherited Anthropic variables
and configures proxy URL, auth, discovery, compaction, and
nonessential-traffic policy. |
| Managed messaging carried a `/v1` API URL and converted it back to a
proxy root while also setting a legacy endpoint variable. | Managed
messaging receives the loopback-safe proxy root and uses the supported
`ANTHROPIC_BASE_URL` contract directly. |
| Managed and interactive policy could diverge while smoke drivers
duplicated both shapes. | `fcc-claude`, messaging, and Claude smoke
drivers use the same canonical envi…
Give messaging clear exact subtree semantics (#1072)
## Problem
Messaging `/clear` did not follow the selected platform message
boundary. Reply clears preserved the selected user prompt, while
standalone clears preserved user-authored messages and could reset FCC
state outside the invoking chat.
## Changes
| Before | After |
| --- | --- |
| Reply `/clear` removed a logical conversation branch but retained the
selected message. | Reply `/clear` deletes the selected message and its
literal reply subtree, including the clear command. |
| Standalone `/clear` retained user prompts and voice notes while
resetting global messaging state. | Standalone `/clear` deletes every
tracked message and resets FCC state only in the invoking platform and
chat. |
| Trees recorded only logical execution parentage. | Trees separately
persist logical execution ancestry and exact prompt/status reply
ownership. |
| Clear coordination used one global admission boundary. | Per-chat
clear generations coordinate admission, voice cancellation, persistence,
and bes…
## Problem
Sixteen OpenAI-compatible providers were represented by
configuration-only subclasses and one factory function each. The sole
provider family also lived under a multi-transport namespace that no
longer described the codebase, while provider defaults, IDs, and
instance caching passed through redundant forwarding layers.
## Changes
| Before | After |
| --- | --- |
| Sixteen provider IDs used configuration-only subclasses. | Immutable
OpenAI-chat profiles configure one concrete provider while preserving
each provider's request policy. |
| Every provider required a dedicated factory function. | Generic
profile construction is the default; only eight adapters with real state
or algorithms retain factories. |
| The sole provider family lived under `providers/transports/`. |
`providers/openai_chat/` directly owns shared request, stream, recovery,
tool, and usage behavior. |
| Provider defaults and IDs passed through forwarding modules, and
constructors repeated fallback resoluti…
Preserve user messages during messaging clear (#1068)
## Problem
Messaging `/clear` used one untyped collection for both internal reply
references and platform deletion targets. Clearing a branch or
cancelling a voice task could therefore delete the customer's prompt or
voice note along with FCC's own status and reply messages.
## Changes
| Before | After |
| --- | --- |
| Tree transitions exposed one message-ID set for repository unindexing
and platform deletion. | Tree transitions now separate internal
`reference_ids` from FCC-owned `clearable_message_ids`. |
| Reply and global clear deleted user prompts and voice notes with FCC
output. | Clear removes FCC statuses, replies, notices, and the explicit
`/clear` command while preserving user-authored messages. |
| The persisted message log accepted ordinary inbound content. | The
clearable-message log accepts only FCC output and explicit clear
commands, and drops legacy user-content entries when loading. |
| Tests treated user-message deletion as successful cleanup. |
Deterministic and …
## Problem
Telegram's online notice was sent directly by the SDK runtime and its
message ID was discarded, so `/clear` could not delete it. Moving
delivery into the workflow also needs to keep slow sends from blocking
commands and prevent acknowledged notices from losing clear ownership.
## Changes
| Before | After |
| --- | --- |
| The Telegram runtime sent a transport-specific startup side effect. |
The platform declares a semantic notice intent that the application
gives to the workflow after transport readiness. |
| Startup delivery bypassed the persisted message log. | The workflow
renders and records each acknowledged notice in the same bounded log
used by `/clear`. |
| Serializing send and record held workflow state across platform I/O. |
A dedicated clear generation reserves publication, delivery runs outside
the state lock, and a short receipt finalizer commits or compensates. |
| Concurrent clear, cancellation, or record failure could leave a
delivered notice unowned. | Cl…
Use task status as the sole stop confirmation (#1065)
## Problem
Messaging `/stop` edited each affected task status to `Stopped` and also
posted a second success message. The duplicate confirmation added noise
even though the existing status already represented the terminal result.
## Changes
| Before | After |
| --- | --- |
| Successful active, queued, global, and bound-voice stops posted a
second confirmation. | Successful stops use the affected task status as
their sole success UI. |
| Stop commands returned an ambiguous integer or `None`. | A typed
`StopOutcome` carries the cancelled count and terminal status ownership.
|
| Statusless voice cancellation could become silent if confirmations
were removed unconditionally. | Statusless voice cancellation receives
one fallback confirmation. |
| A global stop could under-report work when any affected status was in
another chat. | The invoking chat receives one summary whenever any
affected status is outside its scope. |
| Zero-work global stops reported that zero requests were cancelled.…
Make application shutdown completion-driven (#1056)
## Problem\n\nShutdown could report success after bounded messaging
cleanup, hidden persistence failures, or failed managed-process stops.
An Admin restart could then construct a replacement while the old
runtime still owned work.\n\n## Changes\n\n| Before | After |\n| --- |
--- |\n| Runtime reused bounded interactive stop semantics for terminal
messaging cleanup. | Workflow close cancels work, stops managed
sessions, awaits every claim and recovery task, then flushes
persistence. |\n| Explicit persistence failures were logged and treated
as successful writes. | Explicit flushes and authoritative writes
propagate failure and stay dirty for retry; timer writes remain best
effort. |\n| Managed sessions and aliases were removed before subprocess
termination was confirmed. | Manager and session terminal states prevent
reuse, retain failed owners and PIDs, reject ID collisions, and retry
exact sessions. |\n| Admin restart followed the restart request even
after incomplete shutdown. | Superv…
Canonicalize provider failure and retry ownership (#1046)
## Problem
Provider SDK classification, retry policy, canonical failures, and
downstream wire errors shared exception types across layers. That
blurred ownership and let cleanup or provisional Responses tool failures
mask the real provider diagnostic.
## Changes
| Before | After |
| --- | --- |
| Provider failures carried Anthropic wire types and core code
classified OpenAI/httpx errors. | Protocol-neutral `ExecutionFailure`
values cross layers, providers classify SDK errors, and protocol
packages map wire types. |
| Provider adapters could author terminal wire events. | The HTTP commit
boundary selects non-2xx JSON or a protocol terminal event with one
ingress request ID. |
| Retry policy and diagnostic handling were spread across core and
provider modules. | Providers own the unchanged retry budgets while
neutral core utilities own bounded credential redaction. |
| Stream cleanup could replace an already-mapped provider failure. |
Cleanup records safe metadata and preserves the ca…
Introduce a typed application boundary for provider execution (#1045)
## Problem
The HTTP adapter owned model routing, provider execution, and
runtime-facing contracts, so API handlers depended on provider
implementation types. Provider preflight also discovered private request
builders dynamically, obscuring the boundary that must fail before
streaming begins.
## Changes
| Before | After |
| --- | --- |
| `api/` owned model routing and shared provider execution. |
`application/` owns routing and a settings-independent
`ProviderExecutor`. |
| API handlers accepted `BaseProvider` callbacks. | API handlers consume
the narrow structural `ProviderPort`. |
| `BaseProvider` discovered `_build_request_body` dynamically. | Both
transport families implement explicit abstract preflight, with LM Studio
composing context validation. |
| Request leases, task control, and provider model metadata had
adapter/provider owners. | Application-owned ports and immutable values
define those cross-package contracts. |
| Boundary direction was implicit. | Architecture contra…
Replace global runtime resources with explicit ownership (#1042)
## Problem
Provider, messaging, and transcription resources relied on
process-global state, leaving replacement, cancellation, and shutdown
ownership ambiguous. Separate server lifetimes could share
event-loop-bound resources or retain failed cleanup work.
## Changes
| Before | After |
| --- | --- |
| Provider clients found limiters through global singleton and scoped
registries. | Each provider instance receives and owns one explicitly
constructed limiter. |
| Messaging queues and voice pipelines relied on singleton or
module-global state. | Each platform owns its limiter and outbox, while
the application owns one injected transcriber. |
| Messaging shutdown mixed ingress, active work, delivery, and SDK
cleanup. | Application shutdown quiesces ingress, drains work, closes
delivery, then releases transcription and providers. |
| Cancelled or failed provider cleanup could be forgotten or treated as
complete. | The provider manager retains shielded generation and
unpublished-runtime c…
Establish single-owner runtime with stream-safe provider hot swaps (#1036)
## Problem
Provider runtime ownership was split between lifecycle code and mutable
FastAPI state, so Admin replacements could leak the new runtime,
double-close the old runtime, or close providers still serving active
streams. The API package also owned concrete process composition,
obscuring subsystem boundaries.
## Changes
| Before | After |
| --- | --- |
| FastAPI routes inspected several concrete `app.state` resources. |
FastAPI receives one explicit `ApiServices` boundary and stores only
`app.state.services`. |
| Admin Apply persisted config and directly replaced one runtime
reference. | Admin Apply validates a candidate, commits atomically, and
publishes it through the single runtime owner. |
| Provider replacement could close clients used by active streams. |
Generation leases retain old providers until each streaming or
non-streaming response finishes. |
| Provider generations owned discovery state and model metadata. |
`ProviderRuntimeManager` owns one application-lifetime …