[pull] main from CodebuffAI:main

.github/workflows/ci.yml

@@ -135,6 +135,10 @@ jobs:
       - name: Build SDK before tests
         run: cd sdk && bun run build
 
+      - name: Prebuild CLI agents
+        if: matrix.package == 'cli'
+        run: cd cli && bun run prebuild:agents
+
       - name: Run ${{ matrix.package }} tests
         uses: nick-fields/retry@v3
         with:
@@ -228,6 +232,10 @@ jobs:
       - name: Build SDK before integration tests
         run: cd sdk && bun run build
 
+      - name: Prebuild CLI agents
+        if: matrix.package == 'cli'
+        run: cd cli && bun run prebuild:agents
+
       - name: Run ${{ matrix.package }} integration tests
         uses: nick-fields/retry@v3
         with:

evals/buffbench/main.ts

@@ -8,7 +8,7 @@ async function main() {
   // Use 'external:codex' for OpenAI Codex CLI
   await runBuffBench({
     evalDataPaths: [path.join(__dirname, 'eval-codebuff.json')],
-    agents: ['base-deep-evals'],
+    agents: ['base2-free'],
     taskConcurrency: 5,
   })
 

web/src/app/api/v1/chat/completions/_post.ts

@@ -368,8 +368,8 @@ export async function postChatCompletions(params: {
       if (bodyStream) {
         // Streaming request — route to SiliconFlow/CanopyWave/Fireworks for supported models
         const useSiliconFlow = false // isSiliconFlowModel(typedBody.model)
-        const useCanopyWave = isCanopyWaveModel(typedBody.model)
-        const useFireworks = false // isFireworksModel(typedBody.model)
+        const useCanopyWave = false // isCanopyWaveModel(typedBody.model)
+        const useFireworks = isFireworksModel(typedBody.model)
         const stream = useSiliconFlow
           ? await handleSiliconFlowStream({
               body: typedBody,