You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: AGENTS.md
+1-2Lines changed: 1 addition & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -169,8 +169,7 @@ OpenBrowser uses Jinja2 templates for agent prompts, enabling dynamic content in
169
169
- Model profile is resolved from session metadata and exposed to prompt rendering as `model_profile` / `small_model`; see `server/agent/manager.py` and `server/agent/tools/prompt_context.py`
170
170
- Tool prompt variants are split by model profile under `server/agent/prompts/small_model/` and `server/agent/prompts/big_model/`
171
171
- Small-model browser guidance intentionally avoids `keywords` fallback and leans harder on same-mode highlight pagination when dense UI may be split across collision-aware pages
172
-
- Observation rendering also differs by model profile: large models keep clickable highlights compact (`... and N clickable elements`), while small models include clickable element HTML in the LLM-visible observation text for extra semantic grounding
173
-
- The small-model clickable-observation branch is implemented in `server/agent/tools/base.py`; the per-conversation `small_model` flag is attached in `server/agent/tools/browser_executor.py`
172
+
- Observation rendering now includes clickable element HTML for all model profiles; prompt differences still vary by model profile
174
173
175
174
### Keyword Discipline
176
175
- Highlight pagination remains the default discovery flow for controls and dense UI
0 commit comments