[pull] main from Alishahryar1:main - #255
Merged
Merged
Conversation
## Why Alibaba Cloud Model Studio pay-as-you-go keys need their own provider configuration and model discovery. The existing QwenCloud Token Plan and Coding Plan entries use subscription-specific endpoints. ## How - Add an Alibaba Cloud provider card with API key, base URL, and proxy settings. Default to Singapore and allow regional or workspace endpoints with a matching regional key. - Reuse FCC's OpenAI Chat Completions transport for Anthropic Messages and OpenAI Responses clients, including streaming, function calls, and thinking output. - Discover tool-capable inference models through the paginated DashScope `/api/v1/models` endpoint. Require advertised chat-compatible capabilities and exclude Realtime models before publishing model IDs. Read model context limits, output limits, and input modalities. - Validate model IDs across every page, including excluded models. Reject duplicate IDs and inconsistent catalogs so pagination cannot silently publish a partial model list. - Leave thinking computation at the model's upstream defaults. Thinking on/off, effort, and budget controls are not translated in this integration. - Document setup and add the provider to the existing smoke configuration. <!-- greptile_comment --> <!-- greptile_summary --> <h2><a href="https://app.greptile.com/api/retrigger?id=71089453"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img alt="Retrigger" src="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2" align="right"></picture></a>Confidence Score: 5/5</h2> <!-- greptile-risk --> No outstanding findings block merging. <details open><summary>Summary</summary> The PR adds an Alibaba Cloud provider with model discovery and chat transport. Subsequent changes filter the model catalog by capability and reject duplicate model IDs. No outstanding findings remain. </details> <!-- greptile_confidence_score:5 --> <sub>Reviews (2) · Last reviewed commit: ["Fix Alibaba model compatibility and pagi..."](https://github.com/alishahryar1/free-claude-code/commit/fcc8d505e7aead346d8033fe1d66fe79da1f1afa)</sub> <!-- /greptile_comment -->
![https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source](https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source){kind=link}
![https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img](https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img){kind=link}
## Why Code and messaging open and close a SQLite connection for every transaction. Repeated connection setup and WAL cleanup add storage latency, especially when concurrent writes contend. Three alternating local benchmark rounds at `684884e5` against `7c5d5468`, using eight concurrent callers and WAL/FULL durability, measured these median results: | Workload | Before | After | Throughput improvement | | --- | ---: | ---: | ---: | | Mixed SQL reads/writes | 764 ops/s | 6,608 ops/s | 8.65x | | Code create/read/delete/check cycle | 114 cycles/s | 735 cycles/s | 6.47x | | Messaging tree/message persistence and history restoration | 104 cycles/s | 706 cycles/s | 6.79x | Code-cycle p99 fell from 565 ms to 12 ms; messaging-cycle p99 fell from 569 ms to 15 ms. All 18 trials passed data and integrity checks. These are local storage measurements, not whole-app speedups or proof of the earlier Windows CI timeout's cause. ## How - Let the application-owned `SQLiteDatabase` reuse one writer and four reader connections through standalone SQLAlchemy QueuePool. Match async admission to those capacities so waiting writes leave AnyIO workers available for reads and other work. - Keep each lease inside one synchronous worker. Route legacy import's sequential transactions through its single writer admission slot. - Explicitly roll back failed SQL transactions and discard their connections even when rollback succeeds. A failed callback's traceback can retain a SELECT cursor and its old snapshot, so the next borrower must receive a fresh connection. Successful transactions reuse their connections. - Drain accepted queued/running work, dispose both pools, then release process ownership. Preserve committed-result delivery through caller cancellation and clean up partial startup. Accepted calls retain the existing wait-until-complete behavior. - Cover reuse, cross-thread leases, capacity, reader progress, cancellation, failed transactions, retained-cursor isolation after history decoding fails, import rollback, startup and disposal with deterministic tests. Add the base SQLAlchemy dependency and lock it to 2.1.1. Full local `scripts/ci.ps1` validation passed, including formatting, lint, typing, Python tests and Playwright. Benchmark scripts and raw databases remain in the local untracked investigation folder. <!-- greptile_comment --> <!-- greptile_summary --> <h2><a href="https://app.greptile.com/api/retrigger?id=71137019"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img alt="Retrigger" src="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2" align="right"></picture></a>Confidence Score: 5/5</h2> <!-- greptile-risk --> No identified finding blocks merging. <details open><summary>Summary</summary> The PR pools SQLite connections for Code and messaging. The latest commit discards connections after failed transactions and adds regression coverage for retained read cursors and import rollback. No new actionable issue was identified. </details> <!-- greptile_confidence_score:5 --> <sub>Reviews (2) · Last reviewed commit: ["Discard SQLite connections after failed ..."](https://github.com/alishahryar1/free-claude-code/commit/234b24714a52b68e942ed79f9ac3bf53e7d59a36)</sub> <!-- /greptile_comment -->
## Why Server startup emptied the log before opening its append-mode sink, erasing the previous run's unrotated diagnostics. ## How Open the existing log in append mode during initial setup and forced reconfiguration. Keep the `LOG_FILE` override, which smoke tests use to collect diagnostics, with `~/.fcc/logs/server.log` as the default. Regression tests verify that records survive separate process launches and forced logging reconfiguration. <!-- greptile_comment --> <!-- greptile_summary --> <h2><a href="https://app.greptile.com/api/retrigger?id=71158867"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img alt="Retrigger" src="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2" align="right"></picture></a>Confidence Score: 5/5</h2> <!-- greptile-risk --> Safe to merge based on the logging behavior checked. What we checked: - Ran two-process logging restart reproduction scripts before and after applying the PR, and confirmed that the PR version preserves the first run and appends the second run, whereas the pre-PR version retains only the second run. <a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="16" align="absmiddle"></a> - Executed forced-reconfiguration reproduction scripts for base and head; observed that the pre-PR run lost 1,001 records while the PR run retained all 2,001 records; focused logging regression tests also passed. <a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="16" align="absmiddle"></a> - Reviewed the logging configuration changes, noting that the truncating write was removed and the file sink now uses append mode, which explains why the PR run includes both first and second runs. <a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="16" align="absmiddle"></a> - Validated that the 2,001-record scenario with PR #1940 completes without missing records, confirming full retention under the changed code. <a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="16" align="absmiddle"></a> <details open><summary>Summary</summary> The PR stops clearing the server log during initial setup and forced reconfiguration, preserving records across runs. It adds regression tests for both cases. No merge-blocking issue was found. </details> <!-- greptile_confidence_score:5 --> <sub>Reviews (1) · Last reviewed commit: ["Keep LOG\_FILE override and existing smok..."](https://github.com/alishahryar1/free-claude-code/commit/997f472662a6d9d51384f0b6dfc5985c5aae554c)</sub> <!-- /greptile_comment -->
## Why FCC's custom JSON formatter dropped exception details and bound fields outside its fixed allowlist. An intercepted exception could be written as only "request failed", losing the information needed to diagnose it. ## How Use Loguru's built-in JSON serialization (`serialize=True`) and delete the custom formatter and its redaction functions. New JSON lines contain Loguru's `text` and `record` fields, including rendered tracebacks, exception metadata and bound/contextual fields. The sink preserves supplied message text without an additional redaction hook, as agreed for these local diagnostic logs. Update trace readers and smoke assertions to read `record.extra.trace_payload` and log readers to handle Loguru's UTF-8 output. <!-- greptile_comment --> <!-- greptile_summary --> <h2><a href="https://app.greptile.com/api/retrigger?id=71172606"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/RetriggerDark.svg?v=2"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2"><img alt="Retrigger" src="https://greptile-static-assets.s3.amazonaws.com/badges/Retrigger.svg?v=2" align="right"></picture></a>Confidence Score: 5/5</h2> <!-- greptile-risk --> No outstanding findings block merging. <details open><summary>Summary</summary> The PR updates serialized trace logging and its readers. The smoke trace parser test now uses fixed output instead of reconfiguring process-wide logging. </details> <!-- greptile_confidence_score:5 --> <sub>Reviews (2) · Last reviewed commit: ["Isolate smoke trace parser test from glo..."](https://github.com/alishahryar1/free-claude-code/commit/3fb4fd6484775b4fb83d58f5fc9616754717810e)</sub> <!-- /greptile_comment -->
## Why Uvicorn's console handlers stop server errors, access logs and the Admin UI startup message from reaching FCC's log file. Desktop startup can also inherit console handlers from an earlier Uvicorn configuration. The restart browser test can reply to already-aborted status requests on slow runners, leaving the UI waiting indefinitely for the mocked new server. ## How Attach FCC's existing Loguru interceptor to Uvicorn's server and access loggers. Preserve Uvicorn's console handlers for CLI startup. Remove console handler and formatter definitions for desktop startup, allowing Windows GUI launch with no stdout or stderr. Copy Uvicorn's configuration so its global defaults remain unchanged. The file receives each record once, preserves exception tracebacks and respects the configured file log level across server restarts. Make the restart browser mock answer every status request immediately with the current simulated server state. Wait for the matching response before checking each intermediate state, preserving coverage without holding requests past their timeout.
pull
Bot
merged commit Sep 29, 2026
9e23905
into
Co-Contribute-OpenSource:main
10 of 11 checks passed
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )