i99er/ai_mirrors
AI MIRRORS
Owner-controlled, Telegram-first AI workspace for personal research, prototyping and recoverable AI-assisted workflows.
AI MIRRORS is not a public chatbot service. It is a private experimental system that connects model routing, bounded tools, generated artifacts, application prototypes and state recovery behind one owner interface.
Start here
- Public product page:
/ - Liveness:
/health - Readiness:
/ready - Owner guide: `OWNER_GUIDE_RU.md`
- App engine: `APP_V2_README.md`
- Repository map: `REPOSITORY_MAP.md`
- Runtime build policy: `RUNTIME_BUNDLE_ALLOWLIST.csv`
What the repository contains
Operating principles
- Owner authorization before consequential actions.
- Explicit feature gates and no hidden paid fallback.
- Deterministic offline tests where live services are unnecessary.
- Artifacts, hashes and recovery paths for important work.
- Honest separation between source evidence, offline verification and live behavior.
Current status
Personal, non-commercial prototype. External providers, browser environments and Telegram delivery can fail independently. A passing offline test does not by itself prove live availability.
Most implementation work was produced with AI coding agents under owner-directed requirements, review, testing and release decisions.
Local verification
python -m compileall -q .
python -m pytest -q \
tests/test_service_registry_recovery.py \
tests/test_service_run_execution.py \
tests/test_services_core_dedup.pyThe full release gate is intentionally heavier and runs in the Docker build defined by Dockerfile.
Configuration
Copy .env.example to a local untracked .env and fill only the integrations you intend to use. Never commit credentials or generated runtime state.
Boundaries
AI MIRRORS is not intended for production-critical workloads, high-stakes automated decisions, public model resale or processing confidential information through unreviewed external providers.
Deployment state
DEPLOYMENT_PROVENANCE.json in this snapshot records deployment_authorized: false. That is intentional: the package is prepared for owner review, but deployment must be explicitly authorized and followed by live /health, /ready and Telegram canary checks.
License status
No repository license file is included in this snapshot. Do not describe the source as MIT-licensed or reusable under MIT until the owner deliberately adds a valid LICENSE file.
