CONV-001: Conversation Interface Infrastructure

Mark commented

2026-04-11 03:11:50 +00:00

Owner

Overview

Implement thread-based conversation management for multi-turn interactions.

Current State

Mesh is stateless. Each intent is independent with no context from previous interactions.

Missing Components

Thread Management

Thread ID generation and tracking
Thread lifecycle (create, update, close, archive)
Thread metadata (created_at, last_activity, participant count)

HTTP API

# New endpoints needed:
POST /threads                    # Create new conversation thread
GET /threads/{thread_id}         # Get thread status
POST /threads/{thread_id}/intents # Submit intent within thread context
GET /threads/{thread_id}/history  # Get conversation history
DELETE /threads/{thread_id}      # Close/archive thread

Intent Payload Extension

{
  "type": "write_article",
  "payload": {...},
  "thread_id": "uuid-here",  // NEW: Link to conversation
  "context_window": 5          // NEW: How many previous turns to include
}

Storage

Threads table in ledger
Intent-to-thread mapping
Message sequence ordering

Acceptance Criteria

Thread can be created via API
Intents can reference thread_id
Workers receive thread context
Thread history retrievable
Threads can be closed/archived

EMBED-004: Conversational Memory (depends on this)
CONV-002: Multi-turn Context Management

Priority

HIGH - Required for conversational agents

## Overview Implement thread-based conversation management for multi-turn interactions. ## Current State Mesh is stateless. Each intent is independent with no context from previous interactions. ## Missing Components ### Thread Management - [ ] Thread ID generation and tracking - [ ] Thread lifecycle (create, update, close, archive) - [ ] Thread metadata (created_at, last_activity, participant count) ### HTTP API ```python # New endpoints needed: POST /threads # Create new conversation thread GET /threads/{thread_id} # Get thread status POST /threads/{thread_id}/intents # Submit intent within thread context GET /threads/{thread_id}/history # Get conversation history DELETE /threads/{thread_id} # Close/archive thread ``` ### Intent Payload Extension ```json { "type": "write_article", "payload": {...}, "thread_id": "uuid-here", // NEW: Link to conversation "context_window": 5 // NEW: How many previous turns to include } ``` ### Storage - [ ] Threads table in ledger - [ ] Intent-to-thread mapping - [ ] Message sequence ordering ## Acceptance Criteria - [ ] Thread can be created via API - [ ] Intents can reference thread_id - [ ] Workers receive thread context - [ ] Thread history retrievable - [ ] Threads can be closed/archived ## Related - EMBED-004: Conversational Memory (depends on this) - CONV-002: Multi-turn Context Management ## Priority HIGH - Required for conversational agents

Mark added the

feature-request

label

2026-05-12 23:26:55 +00:00

Mark added

blocked:needs-decision

and removed

feature-request

labels

2026-05-21 02:17:27 +00:00

Mark removed the

blocked:needs-decision

label

2026-05-28 07:19:11 +00:00

Mark added the

triage-approved

label

2026-06-07 05:34:36 +00:00

Mark

2026-06-16 03:05:59 +00:00

closed this issue
removed the
triage-approved
label

Mark commented

2026-06-16 03:34:43 +00:00

Author

Owner

Review Summary: CONV-001 Conversation Interface Infrastructure

Verdict: PARTIAL — data structures only, no integration.

Implemented

Thread and Turn dataclasses in substrate/ledger/models.py:370-472 with to_dict()/from_dict() and JSON metadata handling.

Critical Gaps

No DB schema — threads/turns tables missing from substrate/ledger/schema.py init_db().
No repository — substrate/ledger/repositories/ has no ThreadRepository.
No ledger ops — MeshLedger has zero thread/turn methods (create_thread, get_thread, add_turn, get_thread_history, etc.).
No API endpoints — docker/api_gateway.py lacks POST /threads, GET /threads/{id}, POST /threads/{id}/intents, GET /threads/{id}/history, DELETE /threads/{id}.
Intent payload not extended — submit_intent() does not accept thread_id/context_window.
Workers don't receive context — EphemeralWorker._execute_single_intent() ignores thread context.
No tests — zero coverage in tests/unit/test_ledger.py or elsewhere.

Acceptance Criteria

Criterion	Status
Thread can be created via API	FAIL
Intents can reference thread_id	FAIL
Workers receive thread context	FAIL
Thread history retrievable	FAIL
Threads can be closed/archived	FAIL

Recommended Next Steps

Add threads/turns tables + indexes in schema.py.
Create ThreadRepository in substrate/ledger/repositories/.
Add MeshLedger.create_thread/get_thread/add_turn/get_thread_history/close_thread/archive_thread.
Add the 5 thread endpoints to api_gateway.py.
Extend submit_intent to accept + forward thread_id and context_window.
Update EphemeralWorker to read thread context from env and pass into execution.
Add unit + integration tests for models, repository, API, and worker flow.

The models are well-designed but completely disconnected from persistence, API, and runtime. Closing this as a model-only stub does not satisfy the spec.

## Review Summary: CONV-001 Conversation Interface Infrastructure **Verdict: PARTIAL — data structures only, no integration.** ### Implemented - `Thread` and `Turn` dataclasses in `substrate/ledger/models.py:370-472` with `to_dict()`/`from_dict()` and JSON metadata handling. ### Critical Gaps 1. **No DB schema** — `threads`/`turns` tables missing from `substrate/ledger/schema.py` `init_db()`. 2. **No repository** — `substrate/ledger/repositories/` has no `ThreadRepository`. 3. **No ledger ops** — `MeshLedger` has zero thread/turn methods (`create_thread`, `get_thread`, `add_turn`, `get_thread_history`, etc.). 4. **No API endpoints** — `docker/api_gateway.py` lacks `POST /threads`, `GET /threads/{id}`, `POST /threads/{id}/intents`, `GET /threads/{id}/history`, `DELETE /threads/{id}`. 5. **Intent payload not extended** — `submit_intent()` does not accept `thread_id`/`context_window`. 6. **Workers don't receive context** — `EphemeralWorker._execute_single_intent()` ignores thread context. 7. **No tests** — zero coverage in `tests/unit/test_ledger.py` or elsewhere. ### Acceptance Criteria | Criterion | Status | |---|---| | Thread can be created via API | FAIL | | Intents can reference thread_id | FAIL | | Workers receive thread context | FAIL | | Thread history retrievable | FAIL | | Threads can be closed/archived | FAIL | ### Recommended Next Steps 1. Add `threads`/`turns` tables + indexes in `schema.py`. 2. Create `ThreadRepository` in `substrate/ledger/repositories/`. 3. Add `MeshLedger.create_thread/get_thread/add_turn/get_thread_history/close_thread/archive_thread`. 4. Add the 5 thread endpoints to `api_gateway.py`. 5. Extend `submit_intent` to accept + forward `thread_id` and `context_window`. 6. Update `EphemeralWorker` to read thread context from env and pass into execution. 7. Add unit + integration tests for models, repository, API, and worker flow. The models are well-designed but completely disconnected from persistence, API, and runtime. Closing this as a model-only stub does not satisfy the spec.

Mark referenced this issue

2026-07-15 11:38:32 +00:00

Decompose: CONV-003: Web Chat Interface #133

Rows
Columns

CONV-001: Conversation Interface Infrastructure #131

Overview

Current State

Missing Components

Thread Management

HTTP API

Intent Payload Extension

Storage

Acceptance Criteria

Related

Priority

Review Summary: CONV-001 Conversation Interface Infrastructure

Implemented

Critical Gaps

Acceptance Criteria

Recommended Next Steps