v2.0.5: FAST_MODEL routing + tool-use summaries + Qwen dispatch + Arena

Source-level recon of QwenLM/qwen-code (Apache-2.0) informed 4 lifts:

1. FAST_MODEL config: optional env var routes cheap LLM calls (titles,
   summaries, labeling) to a smaller model on llama-swap. auto_name.ts
   uses ctx.config.FAST_MODEL ?? session.model. Set FAST_MODEL=nemotron-
   nano-4b to avoid loading the 35B model for 20-token title generation.

2. Tool-use summaries (services/inference/tool-summaries.ts): utility
   that generates "git-commit-subject-style" labels for tool batches via
   a fast-model LLM call. System prompt + truncation logic ported from
   Qwen Code's toolUseSummary.ts. Exported via @boocode/server/inference
   for BooCoder's dispatcher to call after task completion.

3. Qwen as dispatchable agent: added to agent-probe.ts KNOWN_AGENTS.
   PTY dispatch builds: qwen -p "<task>" --output-format stream-json
   (NDJSON structured events over stdout). Env: OPENAI_BASE_URL +
   OPENAI_API_KEY points Qwen Code at llama-swap. execution_path CHECK
   constraint extended with 'qwen'.

4. Arena routes (routes/arena.ts): POST /api/arena dispatches the same
   task to N contestants (2-5, each with different agent/model), each
   getting its own task row linked by arena_id UUID. GET /api/arena/:id
   shows all contestants. POST /api/arena/:id/select/:task_id marks
   winner. Schema: arena_id column added to tasks.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-25 14:05:59 +00:00
parent 06116f31b3
commit e423579e99
10 changed files with 235 additions and 2 deletions

View File

@@ -7,6 +7,7 @@ const KNOWN_AGENTS: Array<{ name: string; supportsAcp: boolean }> = [
{ name: 'goose', supportsAcp: true },
{ name: 'claude', supportsAcp: false },
{ name: 'pi', supportsAcp: false },
{ name: 'qwen', supportsAcp: false },
];
/**

View File

@@ -8,6 +8,7 @@
* Supported agents:
* - claude: `claude -p --model <model>` (print mode, reads task from stdin)
* - opencode: `echo <task> | opencode` (stdin pipe — exact flags TBD)
* - qwen: `qwen -p <task> --output-format stream-json` (NDJSON structured output)
* - goose: stub (not yet supported)
* - pi: stub (not yet supported)
*/
@@ -51,6 +52,12 @@ function buildAgentCommand(agent: string, task: string, model?: string): string
? `echo '${escapedTask}' | opencode --model '${model}'`
: `echo '${escapedTask}' | opencode`;
case 'qwen':
// Qwen Code: structured JSON output mode for parseable events
return model
? `qwen -p '${escapedTask}' --model '${model}' --output-format stream-json`
: `qwen -p '${escapedTask}' --output-format stream-json`;
case 'goose':
// Not yet verified for non-interactive use
return null;