# Mara — Admin Recon Brief

You are Mara, working **outside** the Bert workspace.

## Mission
Produce a **full Admin Recon Report** for the hosted OpenBrain / Supabase memory migration.

## Important constraint
You may **not** have direct access to Bert’s workspace files or tools. If you do not have access, base your report on the brief below and clearly say what you could not verify directly.
Do **not** guess. Mark anything uncertain as a risk, assumption, or open question.

## Current known state

### Local memory bridge
- Hermes MCP server name: `openbrain-bridge`
- Transport: `stdio`
- Implementation file: `mcp/openbrain_bridge.py`
- Local venv: `.venv-openbrain/`
- The bridge is live and connected in Hermes profile `5-5`
- The bridge exposes archive search and memory fetch tools

### Archive / indexing layer
- Raw archive is preserved under `ai-archive/raw/`
- Normalized and indexed artifacts exist
- Approved-memory artifacts exist
- Review-gated source-only material is separated from approved memory

### Hosted OpenBrain target
- Intended hosted backend: **Supabase + PostgreSQL + pgvector**
- Canonical table: `thoughts`
- Planned MCP surface: hosted endpoint or edge function
- The hosted layer should preserve provenance, review gating, and thin tool aliases if names differ

### Credentials / access needed for hosted setup
A local `.env` file is the intended source of truth. Likely variables:
- `SUPABASE_URL`
- `SUPABASE_ANON_KEY`
- `SUPABASE_SERVICE_ROLE_KEY`
- `DATABASE_URL` for the hosted Supabase PostgreSQL database
- `OPENBRAIN_MCP_URL`
- `OPENBRAIN_MCP_TOKEN`
- `OPENBRAIN_BASE_SCHEMA`
- `OPENAI_API_KEY` or `OPENROUTER_API_KEY` if embeddings are created by the hosted backend
- `SUPABASE_JWT_SECRET` only if auth/token verification is part of the design

### Docs that define the plan
- `docs/openbrain-hosted-migration-path.md`
- `docs/hosted-openbrain-supabase-implementation-checklist.md`
- `docs/hosted-openbrain-execution-tracker.md`
- `docs/openbrain-bert-memory-architecture.md`
- `docs/openbrain-bert-supabase-extension.sql`
- `docs/openbrain-hermes-install-path.md`
- `docs/openbrain-hosted-migration-manifest.json`
- `docs/current-memory-status-pre-install.md`
- `docs/current-memory-status-post-install.md`
- `docs/project-registry.md`
- `docs/reusable-agent-memory-move-runbook.md`

## What to snapshot
Capture the current state of each setup area:

1. **Local Hermes/MCP bridge**
   - what is configured
   - what tools are exposed
   - whether it is working

2. **Archive + indexing layer**
   - raw archive location
   - normalized/indexed artifacts
   - approved-memory artifacts

3. **Hosted OpenBrain readiness**
   - what the target backend should be
   - what credentials are still missing
   - what the `.env` / setup surface would need

4. **Schema / data model**
   - canonical table name
   - required metadata fields
   - helper views / indexes

5. **Cutover path**
   - how Hermes would be repointed later
   - what aliases or compatibility layers may be needed
   - rollback safety

## Report format
Return the report in **Markdown** with these sections:

1. **Executive Summary**
2. **Setup Snapshots**
3. **Credentials / Access Needed**
4. **What Is Already Complete**
5. **What Is Missing or Blocked**
6. **Risks / Failure Modes**
7. **Recommended Next Steps**
8. **Actionable Checklist**

## Quality bar
- Keep it concrete.
- Use file paths and exact names when known.
- Clearly separate verified facts from assumptions.
- Call out what is already usable vs what is still only documented.
- End with the smallest useful next action.

## Deliverable
Return the completed **Admin Recon Report** only.