Skip to content

alias8818/enoch-agentic-research-system

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2,437 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Enoch Agentic Research System

Enoch — auditable AI research control plane

Enoch is a control plane for bounded autonomous AI research. It turns idea intake, worker execution, evidence capture, generated reports, and public release gates into explicit state transitions instead of trusting that a model session "probably finished."

Docs · Launch site · Research corpus · Promising signals

Control-plane flow

Current public release posture

Surface Current public fact
Runtime 1.41.94
Corpus artifacts 393 AI-generated research artifacts
Packaging/provenance gate 393/393 pass
Strict claim/evidence audit 393/393 pass
Promising signals 6,381 bounded no-paper signals

Those counts are public-release facts, not scientific-validity claims. Enoch makes the difference visible: operational health, signal preservation, paper-corpus readiness, and public trust posture are separate claims.

Why this exists

Long-running autonomous AI work fails in ways ordinary scripts do not:

  • child processes can continue after an agent session appears idle;
  • worker telemetry can disagree with queue state;
  • stale rows and optimistic dashboards can hide blocked work;
  • evidence scatters across machines and run folders;
  • generated reports can overstate results when claim boundaries are not preserved.

Enoch treats those as control-plane problems. The system keeps queue state, worker truth, pause/maintenance controls, evidence sync, artifact generation, and release gates outside the model conversation.

What is in this repository

  • FastAPI control plane for queue state, project state, publication automation, pause/maintenance controls, and dispatch decisions.
  • Worker gate and preflight checks for process-tree truth, telemetry quiet windows, and safe dispatch boundaries. Older code/config may still say wake_gate; treat that as compatibility naming.
  • Research Facility ledgers for source scanning, candidate generation, dedupe/history comparison, novelty/feasibility scoring, and admission decisions.
  • Dashboard V2 for operator-facing readiness, queues, runs, papers, and evidence-gate state.
  • Evidence sync and artifact writer for run notes, metrics, result summaries, evidence bundles, claim ledgers, and generated reports.
  • Release validators for packaging/provenance, strict claim/evidence auditability, public counts, and generated public surfaces.

System shape

Research Facility source scan
  -> generated research candidates
  -> dedupe / score / admission ledger
  -> control-plane ideas workbench
  -> queue candidate
  -> control-plane dispatch gates
  -> worker preflight + worker gate
  -> agent run with process/telemetry supervision
  -> evidence sync
  -> AI-generated research artifact
  -> packaging/provenance + strict claim/evidence gates
  -> public corpus or promising-signal lane

For current runtime, storage, worker, decision-artifact, automation, and compatibility boundaries, see docs/current-runtime-snapshot.md.

Public outputs

The reports produced by Enoch runs are AI-generated research artifacts, not human-authored or peer-reviewed papers. The maintainer releases the corpus for inspection and critique but does not claim personal authorship of the generated papers, arguments, or prose.

Getting started

For a local developer smoke test, start with docs/quickstart.md.

For a full deployment path, see docs/deployment-guide.md. For individual config fields, start from config.example.json and docs/configuration-reference.md.

Never commit live config files or credentials.

Development

uv run pytest -q
python scripts/validate_versioning.py
python3 scripts/validate_runtime_snapshot_links.py
python3 scripts/validate_runtime_deploy.py --source . --runtime /opt/enoch-control-plane --expected-commit HEAD --summary-only

Documentation map

Security

Before publishing or deploying changes, run secret scans and tests. See SECURITY.md. Public docs and images must not expose private hostnames, internal paths, tokens, live operator state, or stale count anchors.

License

Apache License 2.0. See LICENSE.

About

Agentic research control plane: queue state, worker preflight, wake-gated execution, evidence sync, dashboard, alerts, and AI-generated paper packaging.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors