From 7424fe4f9bd838764c1c5042808ab4e2a77e0e82 Mon Sep 17 00:00:00 2001 From: weeihan Date: Thu, 1 Oct 2026 13:19:41 +0800 Subject: [PATCH] feat(ai): switch to qwen2.5-coder-8k model - Use larger coder model (32.8B params) for better reasoning - Update client default model - Update all AI routes to use qwen2.5-coder-8k:latest - Already installed and available on server Ollama Co-Authored-By: Claude Haiku 4.5 --- app/api/dashboard/ai/risk-flags/route.ts | 4 ++-- app/api/incidents/[id]/ai/rca-draft/route.ts | 4 ++-- app/api/incidents/[id]/ai/triage-suggest/route.ts | 4 ++-- app/api/incidents/ai/quality-check/route.ts | 4 ++-- lib/claude/client.ts | 2 +- 5 files changed, 9 insertions(+), 9 deletions(-) diff --git a/app/api/dashboard/ai/risk-flags/route.ts b/app/api/dashboard/ai/risk-flags/route.ts index a62a848..82067b7 100644 --- a/app/api/dashboard/ai/risk-flags/route.ts +++ b/app/api/dashboard/ai/risk-flags/route.ts @@ -107,7 +107,7 @@ export async function POST() { let res: Awaited> try { res = await client.chat.completions.create({ - model: 'qwen3:14b', + model: 'qwen2.5-coder-8k:latest', max_tokens: 2048, messages: [ { @@ -166,7 +166,7 @@ Flag only zones with rising or elevated risk (at most 5 flags; do not flag healt await withUser(session.sub, async tx => writeAuditLog(tx, 'incidents', session.sub, 'ai_risk_flags', - { flags, summary: input.summary, model: 'qwen3:14b' }) + { flags, summary: input.summary, model: 'qwen2.5-coder-8k:latest' }) ) return NextResponse.json({ flags, summary: input.summary }) diff --git a/app/api/incidents/[id]/ai/rca-draft/route.ts b/app/api/incidents/[id]/ai/rca-draft/route.ts index 16fce36..7e773a0 100644 --- a/app/api/incidents/[id]/ai/rca-draft/route.ts +++ b/app/api/incidents/[id]/ai/rca-draft/route.ts @@ -57,7 +57,7 @@ export async function POST( let res: Awaited> try { res = await client.chat.completions.create({ - model: 'qwen3:14b', + model: 'qwen2.5-coder-8k:latest', max_tokens: 2048, messages: [ { @@ -105,7 +105,7 @@ Provide 3–5 Why steps drilling from immediate cause to root cause. Frame CAPA await withUser(session.sub, async tx => { await writeAuditLog(tx, 'incidents', id, 'ai_rca_draft', { root_cause_summary: draft.root_cause_summary as string, - model: 'qwen3:14b', + model: 'qwen2.5-coder-8k:latest', }) }) diff --git a/app/api/incidents/[id]/ai/triage-suggest/route.ts b/app/api/incidents/[id]/ai/triage-suggest/route.ts index ee26280..b0e090b 100644 --- a/app/api/incidents/[id]/ai/triage-suggest/route.ts +++ b/app/api/incidents/[id]/ai/triage-suggest/route.ts @@ -48,7 +48,7 @@ export async function POST( let res: Awaited> try { res = await client.chat.completions.create({ - model: 'qwen3:14b', + model: 'qwen2.5-coder-8k:latest', max_tokens: 1024, messages: [ { @@ -103,7 +103,7 @@ Output exactly this JSON structure: } await withUser(session.sub, async tx => { - await writeAuditLog(tx, 'incidents', id, 'ai_triage_suggest', { suggestion: input, model: 'qwen3:14b' }) + await writeAuditLog(tx, 'incidents', id, 'ai_triage_suggest', { suggestion: input, model: 'qwen2.5-coder-8k:latest' }) }) return NextResponse.json(input) diff --git a/app/api/incidents/ai/quality-check/route.ts b/app/api/incidents/ai/quality-check/route.ts index b0e44ea..12e7d24 100644 --- a/app/api/incidents/ai/quality-check/route.ts +++ b/app/api/incidents/ai/quality-check/route.ts @@ -40,7 +40,7 @@ export async function POST(request: NextRequest) { let res: Awaited> try { res = await client.chat.completions.create({ - model: 'qwen3:14b', + model: 'qwen2.5-coder-8k:latest', max_tokens: 1024, messages: [ { @@ -87,7 +87,7 @@ Scoring criteria: specificity (location, time, persons involved), completeness ( await withUser(session.sub, async tx => { await writeAuditLog(tx, 'incidents', session.sub, 'ai_quality_check', { - score: input.score, passes: input.passes, model: 'qwen3:14b', + score: input.score, passes: input.passes, model: 'qwen2.5-coder-8k:latest', }) }) diff --git a/lib/claude/client.ts b/lib/claude/client.ts index 2b5281b..7b8d4cc 100644 --- a/lib/claude/client.ts +++ b/lib/claude/client.ts @@ -5,5 +5,5 @@ export function createDeepSeekClient(apiKey: string): OpenAI { } export function createLocalLLMClient(): OpenAI { - return new OpenAI({ apiKey: 'local', baseURL: 'http://localhost:11434/v1' }) + return new OpenAI({ apiKey: 'local', baseURL: 'http://localhost:11434/v1', defaultQuery: { model: 'qwen2.5-coder-8k:latest' } }) }