How padala-varun/memoraeai-1 works
MemoraeAI is a persistent-memory service with a web chat and memory browser. A chat message is stored, then processed for extraction and routed into episodic, semantic, profile, or goal memory; later, retrieved memories are assembled into context for a response. Gmail events enter through the webhook and use the same extraction path. The README documents retrieval, forgetting, decay, and email capabilities; their precise internal wiring is not fully shown in the sampled source, so only documented flows are represented.
Main entry points: main.py API and Workflow · memory_router.py Memory Router · forgetting.py Forgetting and Decay · semantic_pipeline.py Semantic Pipeline
Components 18
User Experience
- Chat and Memories UI
web/js/chat.js
Memory Orchestration
- API and Workflow
app/main.py - Conversation Service
app/user_service.py - Memory Retrieval
app/main.py - Context Builder
app/context_builder.py
Memory Lifecycle
- Memory Detection
app/memory_detector.py - Memory Router
app/memory_router.py - Episodic Pipeline
app/episodic_pipeline.py - Semantic Pipeline
app/semantic_pipeline.py - Profile and Goals
app/profile_goal_pipeline.py - Forgetting and Decay
app/forgetting.py
Stores and Integrations
- Composio Integration
app/composio_integration.py - MongoDB
- Qdrant
- Neo4j
Other components
- User
- Language Model
- Gmail
Connections 28
- User → Chat and Memories UI: sends messages
- Chat and Memories UI → API and Workflow: calls API
- API and Workflow → Conversation Service: stores and reads
- API and Workflow → Memory Detection: extracts memories
- Memory Detection → Language Model: requests extraction
- API and Workflow → Memory Router: routes results
- Memory Router → Episodic Pipeline: dispatches episodic
- Memory Router → Semantic Pipeline: dispatches semantic
- Memory Router → Profile and Goals: dispatches profile and goals
- Memory Router → Semantic Pipeline: forwards profile facts
- Episodic Pipeline → Qdrant: stores and deduplicates
- Semantic Pipeline → Neo4j: resolves fact conflicts
- Semantic Pipeline → Qdrant: stores embeddings
- Profile and Goals → MongoDB: updates profiles and goals
- API and Workflow → Memory Retrieval: retrieves memories
- Memory Retrieval → Qdrant: searches memories
- Memory Retrieval → Neo4j: reads relationships
- Memory Retrieval → MongoDB: reads profile and goals
- API and Workflow → Forgetting and Decay: handles forget requests
- Forgetting and Decay → Qdrant: removes or invalidates
- Forgetting and Decay → Neo4j: removes or invalidates
- Forgetting and Decay → MongoDB: removes or invalidates
- API and Workflow → Context Builder: builds response context
- Context Builder → Language Model: provides context
- Language Model → API and Workflow: returns response
- Gmail → Composio Integration: sends trigger
- Composio Integration → API and Workflow: posts webhook
- API and Workflow → Memory Detection: processes email content