Spaces:
Sleeping
Sleeping
Commit History
-
Split the classifier into a core agent + conditional follow-up agents
The old call_classification_llm was one ~6,500-token prompt doing 8 unrelated sub-tasks every turn regardless of relevance, including full opinion-clash keyword lists for all 5 characters even though only 1 is ever active. New shape: Agent 1 (core: topic/response_mode/user-info/topic_for_tale/ needs_news_fetch/needs_coverage_check) and Agent 2a (opinion-trigger, scoped to just the active character's theme list) run in parallel every turn. Agent 2b (news question formulation) and Agent 2c (philosophy-thread matching) fire only when Agent 1's own output calls for them. Fixes a live bug as a side effect: philosophy_thread_index and needs_coverage_check were never forwarded from the old classifier's output into analyze_message's return dict, so the philosophy-thread route always saw index=-1 and never used a real match. Exposes the new split in the web dashboard as Core classifier / Opinion trigger / Secondary tasks dropdowns, replacing the single Classification dropdown. Co-Authored-B…78243e3 -
Add per-step model override dashboard for classification/perception/reply
classify_perception.py and classify_chat_helper.py previously bypassed the Fast/Smart mode toggle entirely, hardcoded to OpenAI models via a raw openai.OpenAI() client. Switches both to the shared litellm-backed client so Anthropic models are reachable, adds an optional model= override threaded from the request body through run_chat_app, and exposes it in the web Settings panel as three dropdowns (Default/Haiku/Sonnet per step) for testing cost/quality tradeoffs independently of the existing Fast/Smart toggle. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>be7b2a8
-
rename sories with tales and make the call more neutral
411c910
-
refactor: systematic rename of all modules + SQL reorganisation
Python renames (prefix-based naming convention): Agent_* dialogue agents → dialogue_maieutic / dialogue_pragmatic (+ helpers) Agent_* classifiers → classify_chat / classify_perception / classify_life_event supabase_*.py → db_socratic / db_pragmatic / db_life_events / db_user *_utils.py → util_loop_hierarchy / util_philosophy_threads / util_story_chapters / util_topic_log / util_cross_character / util_geo / util_push / util_translate / util_retrieve / util_agent / util_llm proactive_pipeline → pipeline_proactive Prompt_building → prompt_building ask_llm_final_prompt → prompt_final Classification_parameters → classify_parameters db6_topic_logger → db_topic_logger Deleted: Agent_Socratic_Dialogue.py (legacy, superseded by dialogue_maieutic.py) Deleted: dialogues_socrates.json (unused) SQL reorganisation:…a5a8164