Credentials stay sealed
Source and memory credentials are written to GCP Secret Manager as pinned versions. The console returns source metadata, never plaintext credentials.
ContextDB Managed Sources · Private Alpha
Bring customer context from your database into agent memory.
Preview mapped records, backfill them through the same authenticated memory API, continue from durable checkpoints, propagate soft deletion, and inspect failures without building a connector worker yourself.
Current scope: Postgres has operated validation, preview,
backfill, incremental sync, checkpoint, retry, and soft-delete behavior.
Supabase uses the same Postgres-protocol adapter; no separate production
proof is claimed. Snowflake has SQL API contract coverage but no real
customer warehouse proof yet.
What ContextDB operates
A source sync should resume after failure instead of starting over.
Validate
TLS · credentials · relation
Preview
PII-safe mapped records
Backfill
bounded stable tuple pages
Checkpoint
timestamp + primary key
Increment
updates · deletion · retry
Every job has a named state
PostgreSQL stores sources, jobs, attempts, records, checkpoints, ingestion identity, and dead-letter entries. Workers claim bounded leases and synthesize worker-loss attempts after expiry.
Memory writes are idempotent
A source record carries a stable source key and version. Restarting a backfill does not create another logical memory for the same version.
Provider status
What is implemented and what is only proved in tests.
| Source | Current state | Evidence |
|---|---|---|
| Postgres | Private Alpha | Live validation, preview, backfill, update, soft-delete, checkpoint, and cleanup proof |
| Supabase | Private Alpha | Uses the Postgres protocol and durable worker contract; no separate production proof claimed |
| Snowflake | Contract-tested | OAuth/PAT SQL API, strict account routing, polling, paging, and partition tests; no real warehouse proof |
| Databricks, BigQuery, MongoDB | Planned | Not implemented and not sold as shipped connectors |
Mapping
Map business records without pretending they are conversations.
A source maps one relation, primary key, user partition, content field, update timestamp, and optional deletion timestamp. Business records remain in the source; ContextDB stores the customer context selected for agent memory.
source: customer_context provider: postgres relation: public.customer_context mapping: primary_key: id user_id: customer_id content: context_text updated_at: updated_at deleted_at: deleted_at cursor: updated_at + primary_key
Failure handling
Inspect a failed record instead of losing an entire sync.
Bounded retries
Transient source and gateway failures retry within a frozen attempt and wall-clock budget.
Dead-letter review
Permanent or exhausted record failures enter a project-scoped DLQ and can be requeued explicitly.
Pause without deleting
Operators can pause future synchronization while preserving source metadata, history, and checkpoints.
Verified behavior
The proofs are bounded and named.
- A 20-record live Postgres proof passed validation, preview, backfill, one update, one soft deletion, checkpoint, pause, and cleanup.
- An isolated 100,000-record run survived forced worker loss at record 25,000 and completed with zero logical duplicates.
- The 100,000-record proof took 1,209.65 seconds. It is restart evidence, not a throughput target or availability result.
- The three-design-partner onboarding gate remains open.
Questions
Managed Sources questions, answered directly.
Does ContextDB replace my database or warehouse?
ContextDB does not replace your system of record. Managed Sources selects mapped context for agent memory and keeps provenance back to the source.
Can I connect Snowflake today?
The private-alpha implementation and contract tests exist, but no real customer warehouse proof has passed. It should be treated as a design-partner integration.
Are credentials visible in the console?
No. Credentials are written to Secret Manager and are not returned after source creation.
Does hard-delete reconciliation exist?
No. Explicit soft-delete propagation exists. Continuous hard-delete reconciliation remains a known limitation.
Bring one customer-context table into a private-alpha project.
Start with a PII-safe preview, a bounded backfill, and an explicit checkpoint.