fix: séparer grounding (qwen2.5vl) et compréhension (gemma4)

Browse Source

- Grounding : qwen2.5vl:7b hardcodé (seul modèle avec bbox_2d précis)
- Compréhension/VLM : gemma4:e4b via RPA_VLM_MODEL (description, identification)
- Ajout think=False + num_predict=200 pour éviter le mode thinking gemma4
- Variable RPA_GROUNDING_MODEL pour override si besoin

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

...

This commit is contained in:

Dom

2026-04-04 18:48:00 +02:00

parent 0bd0fbb8c5

commit c1ce6a3964

1 changed files with 1191 additions and 41 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

1232

agent_v0/server_v1/api_stream.py

File diff suppressed because it is too large Load Diff