Add Qwen Code CLI support #12

konard · 2026-01-10T16:01:40Z

Summary

This PR implements support for Qwen Code CLI (Alibaba's AI coding agent optimized for Qwen3-Coder models) as a new tool option in agent-commander, similar to how Claude Code CLI, Codex CLI, OpenCode CLI, Gemini CLI, and @link-assistant/agent are implemented.

Key Changes

JavaScript Implementation (js/src/tools/qwen.mjs)

Added complete tool configuration for Qwen Code CLI
Supports stream-json output format for real-time NDJSON streaming
Auto-approval mode with --yolo flag (enabled by default)
Session management with --resume and --continue options
Context options: --all-files and --include-directories
Partial messages support with --include-partial-messages
Model aliases: qwen3-coder, coder, gpt-4o, gpt-4, sonnet, opus

Rust Implementation (rust/src/tools/qwen.rs)

Complete Rust implementation matching JavaScript functionality
All same features: model mapping, argument building, NDJSON parsing
Session ID and usage extraction
Error detection with proper error types
Tests in separate file: rust/tests/qwen_tests.rs (24 tests)

Tool Registry Updates

JavaScript: Added qwenTool to tools registry in js/src/tools/index.mjs
Rust: Added QwenTool to tools registry in rust/src/tools/mod.rs
Both registries now support: Claude, Codex, OpenCode, Agent, Gemini, and Qwen

Unit Testing

JavaScript: 23 tests for Qwen tool (total: 140 tests pass)
Rust: 24 tests for Qwen tool (total: 197 tests pass)
Tests cover all functionality: model mapping, argument building, output parsing, session ID extraction, usage statistics, error detection

Manual E2E Testing (.github/workflows/e2e-qwen.yml)

Added manual workflow for E2E testing with Qwen Code CLI
Supports workflow_dispatch for manual trigger with configurable inputs
Tests Qwen Code CLI installation and configuration
Tests agent-commander integration in dry-run mode

Experiment Script (js/experiments/test-qwen-integration.mjs)

Comprehensive local test script for Qwen tool configuration
Tests all tool functions without requiring API authentication
Can be extended with --live flag for authenticated testing

Changeset

Added changeset for minor version bump

Test Results

✅ All tests pass:

JavaScript: 140 tests passed (23 new for Qwen, 22 for Gemini)
Rust: 197 tests passed (24 new for Qwen, 25 for Gemini)
Local CI checks: cargo fmt, cargo clippy, npm test - all pass
GitHub CI: All workflows passing

Test Plan

How to Run Manual E2E Tests

Go to Actions tab → "E2E Tests - Qwen Code CLI" workflow
Click "Run workflow" and configure optional parameters
The workflow will test:
- Qwen Code CLI installation
- Tool configuration validation
- Agent-commander integration (dry-run)

For authenticated local testing:

npm install -g @qwen-code/qwen-code@latest
qwen  # then run /auth to authenticate
node experiments/test-qwen-integration.mjs --live

Fixes #11

🤖 Generated with Claude Code

Adding CLAUDE.md with task information for AI processing. This file will be removed when the task is complete. Issue: #11

Implements support for Qwen Code CLI (Alibaba's AI coding agent optimized for Qwen3-Coder models) as a new tool option in agent-commander. Key features: - Stream JSON format with --output-format stream-json for real-time NDJSON streaming - Auto-approval mode with --yolo flag (enabled by default for autonomous execution) - Session management with --resume and --continue options - Context options: --all-files and --include-directories - Partial messages support with --include-partial-messages - Model aliases: qwen3-coder, coder, gpt-4o Files changed: - js/src/tools/qwen.mjs: New tool configuration - js/src/tools/index.mjs: Tool registry update - js/test/tools.test.mjs: 23 new tests for Qwen tool - README.md: Documentation and examples - js/.changeset/add-qwen-coder-support.md: Changeset for release Fixes #11 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>

This reverts commit da78cc5.

konard · 2026-01-10T16:15:00Z

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

Public pricing estimate: $6.705090 USD
Calculated by Anthropic: $5.143415 USD
Difference: $-1.561675 (-23.29%)
📎 Log file uploaded as GitHub Gist (951KB)
🔗 View complete solution draft log

Now working session is ended, feel free to review and add any feedback on the solution draft.

konard · 2026-01-10T16:29:44Z

We should also check if link-assistant/hive-mind#556 has any useful ideas for that implementation.

Double check we do unit test everything we, can and it is executed in CI/CD. Also please add support for manual e2e tests (meaning it should be separate manually executed CI workflow) using free plan of Qwen Code CLI.

konard · 2026-01-10T16:32:05Z

🤖 AI Work Session Started

Starting automated work session at 2026-01-10T16:32:03.619Z

The PR has been converted to draft mode while work is in progress.

This comment marks the beginning of an AI work session. Please wait working session to finish, and provide your feedback.

- Add .github/workflows/e2e-qwen.yml for manual E2E testing - Supports workflow_dispatch for manual trigger - Tests Qwen Code CLI installation and configuration - Tests agent-commander integration in dry-run mode - Provides clear instructions for authenticated testing - Add js/experiments/test-qwen-integration.mjs - Comprehensive test script for Qwen tool configuration - Tests model mapping, argument building, output parsing - Tests session ID extraction, usage extraction, error detection - Can be run locally without authentication 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>

konard · 2026-01-10T16:39:39Z

🤖 Work Session Complete

Based on the feedback from @konard, I've made the following updates:

1. Reviewed hive-mind PR #556

I studied the implementation in link-assistant/hive-mind#556 for useful ideas. The implementation here follows similar patterns with:

Model mapping functions
Session management support
NDJSON output parsing
Error detection
Usage statistics extraction

2. Verified Unit Tests

All 118 unit tests pass and are executed in CI/CD:

23 tests specifically for the Qwen tool
Tests cover model mapping, argument building, output parsing, session ID extraction, usage extraction, error detection, and capability flags

3. Added Manual E2E Testing Support

New workflow: .github/workflows/e2e-qwen.yml

Manual workflow triggered via workflow_dispatch
Tests Qwen Code CLI installation
Tests tool configuration validation
Tests agent-commander integration (dry-run mode)
Provides clear instructions for authenticated testing with free tier

New experiment script: js/experiments/test-qwen-integration.mjs

29 comprehensive integration tests
Can be run locally without authentication
Supports --live flag for authenticated testing

How to Run Manual E2E Tests

Via GitHub Actions:
- Go to Actions tab → "E2E Tests - Qwen Code CLI"
- Click "Run workflow"
- Configure optional parameters (prompt, working directory, model)

Locally (authenticated):

npm install -g @qwen-code/qwen-code@latest
qwen  # then run /auth for free tier (2000 requests/day)
node experiments/test-qwen-integration.mjs --live

CI Status

✅ All CI checks passing for the latest commit (70e5641)

konard · 2026-01-10T16:39:56Z

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

Public pricing estimate: $4.708896 USD
Calculated by Anthropic: $2.529908 USD
Difference: $-2.178989 (-46.27%)
📎 Log file uploaded as GitHub Gist (590KB)
🔗 View complete solution draft log

Now working session is ended, feel free to review and add any feedback on the solution draft.

konard · 2026-01-15T14:19:06Z

Both JavaScript and Rust versions must be implemented, have all the same tess. Make sure all our tests in Rust are in separate files, not mixed with code, like we do in JavaScript.

Also ensure all changes are correct, consistent and fully meet original requirements.

konard · 2026-01-15T14:19:55Z

🤖 AI Work Session Started

Starting automated work session at 2026-01-15T14:19:53.093Z

The PR has been converted to draft mode while work is in progress.

This comment marks the beginning of an AI work session. Please wait working session to finish, and provide your feedback.

- Add rust/src/tools/qwen.rs with complete Qwen tool implementation - Model mapping with qwen3-coder, coder, gpt-4o, gpt-4, sonnet, opus aliases - Build args for all Qwen Code options (--yolo, --resume, --continue, etc.) - NDJSON output parsing, session ID extraction, usage stats, error detection - QwenTool struct with capability flags matching JS implementation - Update rust/src/tools/mod.rs to register Qwen tool - Add qwen module and exports - Implement Tool trait for QwenTool - Update ToolRegistry, get_tool, list_tools, is_tool_supported - Move all tool tests to separate test files (per feedback): - rust/tests/qwen_tests.rs (24 tests) - rust/tests/claude_tests.rs (17 tests) - rust/tests/codex_tests.rs (12 tests) - rust/tests/opencode_tests.rs (10 tests) - rust/tests/agent_tests.rs (14 tests) - rust/tests/tools_tests.rs (11 tests) Test results: - Rust: 60 unit tests + 88 integration tests = 148 total (all passing) - JavaScript: 118 tests (all passing) Both JS and Rust implementations are consistent with: - Same model mappings - Same capability flags - Same default values - Same function signatures and behavior Co-Authored-By: Claude Opus 4.5 <[email protected]>

Apply automatic formatting via cargo fmt to: - rust/src/tools/qwen.rs: Align comment indentation - rust/tests/agent_tests.rs: Multi-line assert formatting - rust/tests/claude_tests.rs: Multi-line assert formatting - rust/tests/opencode_tests.rs: Multi-line assert formatting Co-Authored-By: Claude Opus 4.5 <[email protected]>

Replace manual Default impl for QwenErrorResult with #[derive(Default)] as suggested by clippy. This is equivalent since all fields have default values (bool defaults to false, Option defaults to None). Co-Authored-By: Claude Opus 4.5 <[email protected]>

- tests/claude_tests.rs: Remove unnecessary reference in comparison - tests/qwen_tests.rs: Replace collect+is_empty with any() for cleaner code Both fixes address clippy warnings that are treated as errors with -Dwarnings. Co-Authored-By: Claude Opus 4.5 <[email protected]>

konard · 2026-01-15T15:08:52Z

⚠️ Solution Draft Finished with Errors

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

Public pricing estimate: $12.351770 USD
Calculated by Anthropic: $0.000000 USD
Difference: $-12.351770 (-100.00%)

Note: The session encountered errors during execution, but some work may have been completed. Please review the changes carefully.

📎 Log file uploaded as Gist (1375KB)
🔗 View complete solution draft log

Now working session is ended, feel free to review and add any feedback on the solution draft.

konard · 2026-01-17T23:08:05Z

🤖 AI Work Session Started

Starting automated work session at 2026-01-17T23:08:03.655Z

The PR has been converted to draft mode while work is in progress.

This comment marks the beginning of an AI work session. Please wait working session to finish, and provide your feedback.

…ntations Both Qwen Code CLI and Gemini CLI are now supported: - JavaScript: qwen.mjs and gemini.mjs tools - Rust: qwen.rs and gemini.rs modules - Updated tool registries to include both tools - Test files updated to include both tools Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

- Added geminiTool to imports - Combined Qwen and Gemini tests properly - All 140 JavaScript tests now pass Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

konard · 2026-01-17T23:18:06Z

✅ Work Session Complete

All requirements from the feedback have been addressed:

1. Both JavaScript and Rust Versions Implemented ✅

JavaScript: Complete implementation in js/src/tools/qwen.mjs with 23 comprehensive tests
Rust: Complete implementation in rust/src/tools/qwen.rs with 24 comprehensive tests
Both implementations have the same features and functionality

2. Rust Tests in Separate Files ✅

All Rust tests are in rust/tests/qwen_tests.rs (separate file)
No inline tests in source code (verified with grep)
Matches the JavaScript testing pattern

3. Changes Are Correct and Consistent ✅

Merged main branch and resolved conflicts with Gemini implementation
Both Qwen and Gemini tools are now supported
Tool registries updated in both JavaScript and Rust
All 140 JavaScript tests pass
All 197 Rust tests pass

4. CI/CD Status ✅

All CI checks passing on latest commit (29bf13e)
Rust: cargo fmt, cargo clippy, cargo test - all pass
JavaScript: npm test, eslint - all pass

Implementation Summary

Files Changed:

js/src/tools/qwen.mjs - JavaScript implementation
js/src/tools/index.mjs - Added Qwen to registry (alongside Gemini)
js/test/tools.test.mjs - 23 tests for Qwen, 22 for Gemini
rust/src/tools/qwen.rs - Rust implementation
rust/src/tools/mod.rs - Added Qwen to registry (alongside Gemini)
rust/tests/qwen_tests.rs - 24 tests for Qwen

Test Coverage:

JavaScript: 140 tests (23 for Qwen, 22 for Gemini)
Rust: 197 tests (24 for Qwen, 25 for Gemini)

The PR is now ready for review!

🤖 Generated with Claude Code

konard · 2026-01-17T23:18:31Z

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

Public pricing estimate: $5.250169 USD
Calculated by Anthropic: $2.514168 USD
Difference: $-2.736001 (-52.11%)
📎 Log file uploaded as Gist (808KB)
🔗 View complete solution draft log

Now working session is ended, feel free to review and add any feedback on the solution draft.

Initial commit with task details

da78cc5

Adding CLAUDE.md with task information for AI processing. This file will be removed when the task is complete. Issue: #11

konard self-assigned this Jan 10, 2026

konard changed the title ~~[WIP] Add Qwen Coder support~~ Add Qwen Code CLI support Jan 10, 2026

Fix changeset formatting (Prettier)

e5d164e

🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <[email protected]>

konard marked this pull request as ready for review January 10, 2026 16:14

Revert "Initial commit with task details"

4492953

This reverts commit da78cc5.

konard marked this pull request as draft January 10, 2026 16:32

konard marked this pull request as ready for review January 10, 2026 16:39

konard marked this pull request as draft January 15, 2026 14:19

konard and others added 4 commits January 15, 2026 15:28

konard marked this pull request as ready for review January 15, 2026 15:08

konard marked this pull request as draft January 17, 2026 23:08

konard and others added 2 commits January 18, 2026 00:09

Fix merge conflicts in JavaScript test file

29bf13e

- Added geminiTool to imports - Combined Qwen and Gemini tests properly - All 140 JavaScript tests now pass Co-Authored-By: Claude Sonnet 4.5 <[email protected]>

konard marked this pull request as ready for review January 17, 2026 23:17

konard merged commit 3fe9021 into main Jan 18, 2026
44 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Qwen Code CLI support #12

Add Qwen Code CLI support #12

Uh oh!

konard commented Jan 10, 2026 •

edited

Loading

Uh oh!

konard commented Jan 10, 2026

Uh oh!

konard commented Jan 10, 2026 •

edited

Loading

Uh oh!

konard commented Jan 10, 2026

Uh oh!

konard commented Jan 10, 2026

Uh oh!

konard commented Jan 10, 2026

Uh oh!

konard commented Jan 15, 2026

Uh oh!

konard commented Jan 15, 2026

Uh oh!

konard commented Jan 15, 2026

Uh oh!

konard commented Jan 17, 2026

Uh oh!

konard commented Jan 17, 2026

Uh oh!

konard commented Jan 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Add Qwen Code CLI support #12

Add Qwen Code CLI support #12

Uh oh!

Conversation

konard commented Jan 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Key Changes

Test Results

Test Plan

How to Run Manual E2E Tests

Uh oh!

konard commented Jan 10, 2026

🤖 Solution Draft Log

Uh oh!

konard commented Jan 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

konard commented Jan 10, 2026

Uh oh!

konard commented Jan 10, 2026

🤖 Work Session Complete

1. Reviewed hive-mind PR #556

2. Verified Unit Tests

3. Added Manual E2E Testing Support

How to Run Manual E2E Tests

CI Status

Uh oh!

konard commented Jan 10, 2026

🤖 Solution Draft Log

Uh oh!

konard commented Jan 15, 2026

Uh oh!

konard commented Jan 15, 2026

Uh oh!

konard commented Jan 15, 2026

⚠️ Solution Draft Finished with Errors

📎 Log file uploaded as Gist (1375KB) 🔗 View complete solution draft log

Uh oh!

konard commented Jan 17, 2026

Uh oh!

konard commented Jan 17, 2026

✅ Work Session Complete

1. Both JavaScript and Rust Versions Implemented ✅

2. Rust Tests in Separate Files ✅

3. Changes Are Correct and Consistent ✅

4. CI/CD Status ✅

Implementation Summary

Uh oh!

konard commented Jan 17, 2026

🤖 Solution Draft Log

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

konard commented Jan 10, 2026 •

edited

Loading

konard commented Jan 10, 2026 •

edited

Loading

📎 Log file uploaded as Gist (1375KB)
🔗 View complete solution draft log