Project intelligence
Project Meetings
Chunk meeting dài, extract candidate theo project, rồi approve output đã review.
Loaded from the configured OpenAI-compatible `/models` endpoint.
Model guide
Project meetings: use qwen3.6:35b for final extraction; use gemma4:26b when you mainly need knowledge summaries.
| Model | Speed | Use for | Avoid for |
|---|---|---|---|
| qwen3.6:35b | Slow | Vietnamese extraction, strict JSON, long project meetings, high-value runs | Quick smoke tests |
| gemma4:26b | Slow | Long summaries, knowledge notes, broad reasoning over messy context | Cheap repeated batch runs |
| gpt-oss:20b | Balanced | Technical chat, troubleshooting, mixed English/Vietnamese notes | Very strict Vietnamese JSON when Qwen is available |
| gemma3:12b | Balanced | Daily extraction, normal chat review, medium documents | Very long meetings with many decisions |
| qwen3.5:9b | Fast | Short Vietnamese chats, task candidates, quick review passes | Complex cross-chunk reasoning |
| gemma4:e4b | Fast | Draft runs, quick classification, cheap first pass | Final approval extraction on important meetings |
| gemma3:4b | Fast | Smoke tests, tiny notes, UI/debug checks | Production extraction quality |
Chunks
0
Candidates
0
Pending
0
Rejected
0
Applied
0
Outputs
0
Project candidates
Approve writes to `project_outputs`; reject keeps the audit trail.
Chưa có candidate. Nhập meeting id rồi bấm Extract dry-run.
Approved project outputs
0 outputs gần nhất.
| Type | Output | Project | Owner | Confidence |
|---|---|---|---|---|
| Chưa có output được approve. | ||||