git push origin fix-mermaid-architecture-diagrams#19
Conversation
|
Thanks so much for taking the time to convert this @Atharva-1512 ! Moving the architecture diagram to Mermaid is a great idea and will definitely make the documentation cleaner and easier to read. However, taking a look at the proposed flowchart logic, there are a few paths that don't quite match the actual execution flow in session.py:
I went ahead and drafted a revised version of the Mermaid syntax that captures these specific execution paths accurately. Feel free to copy this and update the PR with it! flowchart TD
A[generate_next_turn()] --> B{Orchestrator due?}
B -->|Yes| C[Speak or PASS]
C -->|Speak| EndOrch[Append to history & return turn]
C -->|PASS| D
B -->|No| D{_forced_next_agent set?}
D -->|Yes| E[Use forced agent & clear flag] --> L
D -->|No| F{session_type = DYNAMIC?}
F -->|No| K[Fallback: round robin / argumentative] --> L
F -->|Yes| G{@mention in last message?}
G -->|Yes| H[Use mentioned agent] --> L
G -->|No| I[Score agents by expertise]
I --> J{Top score > 0?}
J -->|Yes| Best[Pick best matching agent] --> L
J -->|No| K
L[agent.generate_response()] --> M{response == PASS?}
M -->|Yes| N[Skip turn, return skipped: True]
M -->|No| O[Append to history with timestamp] --> P[Return turn_data]
|
|
@rosspeili I've made the necessary changes. Please review and suggest any further changes necessary. Thank you. |
|
@Atharva-1512 I went ahead and manually updated the PR description to Thanks a ton for your effort and for being so receptive to tweaking the diagram logic. If you're interested in sticking around, feel free to check out the other Thanks again for the contribution! |
|
Thanks a lot for the support and for merging the PR! |
Fixes #12
Converted the ASCII decision flow diagram in ARCHITECTURE.md
to a Mermaid flowchart for better rendering and readability.