42 lines (31 loc) · 1.62 KB

API Client & Streaming

Status: Partial
Package: beep/api/

Purpose

Async HTTP transport to Beep.AI.Server: OpenAI-compatible chat, models, embeddings, RAG, Coding Assistant bootstrap, agent bundles, token check.

Code locations

Module	Role
`client.py`	`BeepAPIClient`, endpoint methods
`client_llm_support.py`	LLM request helpers
`client_workspace_support.py`	Workspace/bootstrap helpers
`client_agent_bundle_support.py`	Bundle import/export HTTP
`payloads.py`	Request body shaping
`streaming.py`	SSE parsing, tool-call sentinels

User surfaces

All commands that call the server (chat, ask, agent, rag, status, …)
REPL model turns via ChatSession

Current behavior

Bearer auth on /v1/* and /ai-middleware/api/*
Retries on 429/503; configurable timeout; BeepAPIError mapping
Streaming chat completions with tool deltas
Health: GET /api/health; optional GET /ai-middleware/api/tokens/check

Enhancement backlog

ID	Enhancement	Priority	Verification
API-1	OpenAPI-derived client stubs or contract tests against server OpenAPI	P2	CI contract job
API-2	Request/response redaction in debug logs (`BEEP_LOG_JSON`)	P1	Log snapshot test
API-3	Connection pooling metrics in `beep diagnostics`	P2	Diagnostics output
API-4	Automatic server version compatibility warning	P1	Status command test
API-5	Batch embeddings helper for large index jobs	P3	Unit test
API-6	Circuit breaker after repeated 5xx	P2	Simulated failure test