# BitterLog

BitterLog is an append-only evidence workbench for agent-run proof: raw traces,
receipts, prompt/output artifacts, annotations, stable digests, and
account-scoped retrieval.

## Public Product Loop

```text
Agent Run -> Raw Event or Bucket Object -> Digest/Timestamp/Receipt -> Annotation/Query -> Reviewable Proof
```

BitterLog is built for AI-native operators who need to preserve what happened
during agent work, recover the exact source record behind a summary or failure,
and make evidence inspectable for later review.

## Source record stays inspectable

Raw traces stay tied to stable digests and timestamps. Projected reads work
from refs and projections, not a rewritten source of truth.

## Trust Boundaries

- Receipts, request context, stdout, stderr, annotations, and deploy receipts
  can connect into reviewable proof for operators.
- API and CLI ingest preserve the original trace first. Parse later.
- ClickHouse-backed search is a projected read model when analytics is enabled,
  not the append-only ledger itself.
- Account-scoped access routes through Bitter account surfaces rather than a
  BitterLog-only login.

## Current Public Status

The public API status currently reports `analytics.enabled: false`. Public copy
should treat search, SQL, and similarity as conditional projected read models
until live status proves them available. Evidence onboarding remains
request-reviewed and should route through the Bitter account bridge or email,
not a marketing-site form endpoint.

## Links

- HTML home: https://bitterlog.com/
- Public source: https://github.com/sheetgenius/bitterlog-marketing
- API index: https://api.bitterlog.com/api/v1
- API status: https://api.bitterlog.com/api/v1/status
- Health and release identity: https://bitterlog.com/up/
- AI crawler index: https://bitterlog.com/llms.txt
- Full AI context: https://bitterlog.com/llms-full.txt