Feature: Semantic project memory via Dakera — recall architectural decisions across sessions

GPT-Pilot builds applications iteratively across multiple conversations. Each new conversation requires re-loading project context from spec files. This proposes Dakera (https://dakera.ai) as semantic memory of past architectural decisions and debugging sessions.

Problem: GPT-Pilot's task breakdown relies on the project spec file, but semantic recall of past debugging sessions is lost between conversations. 'We hit a Pydantic validation error on the user schema — resolved by adding Optional types' gets re-derived from scratch in every session.

Proposed Integration: At session load, recall(project_description + current_task) retrieves semantically relevant prior decisions. At session save, store_memory(task_summary) persists key outcomes.

In pilot/core/project_state/:

from dakera import DakeraClient

class ProjectMemory: def init(self, project_id: str): self._client = DakeraClient( base_url=os.environ.get('DAKERA_URL', 'http://localhost:3300'), api_key=os.environ.get('DAKERA_API_KEY', ''), ) self._agent_id = f'gptpilot-{project_id}'

def recall_for_task(self, task: str, top_k: int = 5) -> list:
    response = self._client.recall(agent_id=self._agent_id, query=task, top_k=top_k)
    return response.memories if response else []

def store_task_outcome(self, task: str, outcome: str) -> None:
    self._client.store_memory(
        agent_id=self._agent_id,
        content=f'Task: {task}\nOutcome: {outcome}',
    )

Setup: docker run -d -p 3300:3300 -e DAKERA_API_KEY=demo ghcr.io/dakera-ai/dakera:latest / pip install dakera

Happy to open a PR.