API Reference

Base URL: http://127.0.0.1:8000

Auth

• POST /auth/register
• body: { "email": "...", "password": "..." }
• response: access_token, user

• POST /auth/login
• body: { "email": "...", "password": "..." }
• response: access_token, user

• POST /auth/logout
• header: Authorization: Bearer <token>

• GET /auth/me
• header optional: Authorization

Plans & billing

• GET /plans
• vraci aktivni plany (free, pro, business)

• POST /billing/subscribe
• header: Authorization: Bearer <token>
• body: { "plan_code": "pro", "reason": "user_upgrade" }

Chat

• POST /chat
• header optional: Authorization
• header optional: X-Guest-Id
• body:
• message (required)
• conversation_id (optional)
• system (optional)
• max_new_tokens, temperature, top_k (optional)

• GET /chat/conversations
• user: Bearer token
• guest: automaticky nebo pres X-Guest-Id

• GET /chat/conversations/{conversation_id}/messages
• vraci messages jen vlastnikovi (user/guest)

Usage & learning status

• GET /usage/me
• vraci denni usage a limity

• GET /learning/status
• vraci posledni autodeploy report
• zdroj: enoa_gpt/checkpoints/autodeploy_last.json

• GET /learning/healthcheck
• lehky operational health endpoint pro learning stack
• vraci:
• prod_checkpoint_exists
• candidate_checkpoint_exists
• autodeploy_report_exists
• autodeploy_report_decision
• autodeploy_report_age_minutes

• GET /metrics
• Prometheus text metrics endpoint
• obsahuje runtime/learning zakladni metriky (model loaded, checkpoint existence, report age)

• GET /ops/recovery-status
• operational endpoint pro watchdog/recovery stav
• vraci:
• recovery_fail_count
• restarts_total
• restarts_last_hour
• last_restart_time
• recent_log_lines

Monitoring auth:

• pokud je nastaveno ENOAI_MONITOR_API_KEY, endpointy
• /learning/status
• /learning/healthcheck
• /metrics
• /ops/recovery-status

vyzaduji header X-Monitor-Key: <value>