API Reference¶

All external interfaces of this system use the /api/v1/ prefix and follow RESTful conventions. Request and response bodies are JSON (file uploads use multipart). This document lists all endpoints grouped by module.

Transport conventions

All endpoints (except health check and read-only ops endpoints) are authenticated via the verify_api_key dependency.
Request header Content-Type: application/json (except file uploads, which use multipart/form-data).
Time fields use ISO8601 (UTC) format, e.g. 2026-07-03T08:30:00+00:00.

Authentication¶

The system authenticates via the X-API-Key request header. The logic is implemented by verify_api_key in app/core/security.py.

Production mode (API_KEY non-empty)Development mode (API_KEY empty)Authentication failure response

curl -H "X-API-Key: $API_KEY" \
     -H "Content-Type: application/json" \
     http://localhost:8000/api/v1/chat \
     -d '{"message": "Hello"}'

# When API_KEY is empty, the system enters no-auth mode; no X-API-Key header required
curl -H "Content-Type: application/json" \
     http://localhost:8000/api/v1/chat \
     -d '{"message": "Hello"}'

Security note

Development mode is for local debugging only. Production must set a non-empty API_KEY; otherwise any caller can access the system.

{
  "detail": "Invalid or missing API key"
}

Authentication failures uniformly return 401 Unauthorized.

Unified Error Response¶

Beyond HTTP status codes, all error responses follow a unified structure:

{
  "detail": "Error description text"
}

Status Code	Meaning	Trigger Scenario
200	Success	Successful GET / POST / PUT
201	Created	Resource-creating endpoints such as creating an experiment
204	No Content	Endpoints with no response body, such as recording experiment metrics
400	Bad Request	Invalid parameter value (e.g., alert level not in enum)
401	Unauthorized	`X-API-Key` missing or invalid
404	Not Found	session_id / doc_id / report_id / trace_id does not exist
409	Conflict	State transition conflicts such as duplicate agent acceptance or sending messages to a non-`assigned` session
422	Validation Failed	Pydantic validation failure (e.g., `min_length=1` constraint)
500	Internal Error	Uncaught server-side exception

Endpoint List Overview¶

Chat Module¶

Endpoint	Method	Description	Auth
`/api/v1/chat`	POST	Synchronous chat; returns a full `ChatResponse`	✅
`/api/v1/chat/stream`	POST	SSE streaming chat; event sequence meta → token → done	✅
`/api/v1/gateway`	POST	Unified multi-channel access gateway	✅

Agent Assist¶

Endpoint	Method	Description	Auth
`/api/v1/agent/sessions/pending`	GET	Pending session list (sorted by priority descending)	✅
`/api/v1/agent/sessions/{session_id}`	GET	Session details (including EscalationCard + history)	✅
`/api/v1/agent/sessions/{session_id}/accept`	POST	Agent accepts the session (CAS pending → assigned)	✅
`/api/v1/agent/sessions/{session_id}/messages`	POST	Agent sends a message appended to history	✅
`/api/v1/agent/sessions/{session_id}/knowledge-recommend`	POST	Knowledge recommendation assist	✅
`/api/v1/agent/sessions/{session_id}/business-assist`	POST	Business query assist (with masking)	✅
`/api/v1/agent/sessions/{session_id}/resolve`	POST	Mark resolved (CAS assigned → resolved)	✅
`/api/v1/agent/sessions/{session_id}/solution`	POST	Consolidate a solution back to the knowledge base	✅

Knowledge Base Management¶

Endpoint	Method	Description	Auth
`/api/v1/knowledge/ingest`	POST	Document ingestion (multipart)	✅
`/api/v1/knowledge/stats`	GET	Knowledge base statistics	✅
`/api/v1/knowledge/documents`	GET	Paginated document list	✅
`/api/v1/knowledge/documents/{doc_id}`	GET	Document details (with version history)	✅
`/api/v1/knowledge/documents/{doc_id}`	DELETE	Delete a document (removes all chunks)	✅
`/api/v1/knowledge/quality/check`	POST	Quality check inspection	✅
`/api/v1/knowledge/documents/{doc_id}/rollback`	POST	Roll back to a specified version	✅
`/api/v1/knowledge/canary/ingest`	POST	Write to the canary collection	✅
`/api/v1/knowledge/canary/compare`	POST	Compare main collection vs canary collection	✅

Document Updates¶

Endpoint	Method	Description	Auth
`/api/v1/update/full`	POST	Full update (monthly)	✅
`/api/v1/update/incremental`	POST	Incremental update (weekly)	✅
`/api/v1/update/file`	POST	Single-file real-time update	✅
`/api/v1/update/status`	GET	Result of the most recent update	✅

Ticket Mining and Retrieval Tuning¶

Endpoint	Method	Description	Auth
`/api/v1/mining/tickets`	POST	Trigger historical ticket knowledge mining	✅
`/api/v1/mining/status`	GET	Most recent mining report	✅
`/api/v1/tuner/params`	GET	Query current tuning parameters	✅
`/api/v1/tuner/params`	PUT	Update tuning parameters (takes effect immediately)	✅
`/api/v1/tuner/reset`	POST	Reset to default parameters	✅

Retrieval Evaluation¶

Endpoint	Method	Description	Auth
`/api/v1/evaluation/run`	POST	Trigger an evaluation run	✅
`/api/v1/evaluation/reports`	GET	Historical report summary list	✅
`/api/v1/evaluation/reports/{report_id}`	GET	Single report details	✅

Performance Monitoring¶

Endpoint	Method	Description	Auth
`/api/v1/performance/metrics`	GET	Comprehensive performance metrics	❌
`/api/v1/performance/cache/stats`	GET	Hot cache statistics	❌
`/api/v1/performance/cache/invalidate`	POST	Clear the hot cache	❌

Performance endpoints are unauthenticated

Performance monitoring endpoints are not authenticated, so ops dashboards and knowledge base update pipelines can call them without credentials. In production, we recommend adding IP allowlists or authentication at the reverse proxy layer.

Observability¶

Endpoint	Method	Description	Auth
`/api/v1/observability/circuit-breakers`	GET	List all circuit breaker states	❌
`/api/v1/observability/circuit-breakers/{name}/reset`	POST	Manually reset a circuit breaker	❌
`/api/v1/observability/alerts`	GET	Query alert list (supports filtering)	❌
`/api/v1/observability/health`	GET	Health check report	❌
`/api/v1/observability/token-usage`	GET	Token usage statistics	❌

Monitoring (Agent dimension)¶

Endpoint	Method	Description	Auth
`/api/v1/monitor/overview`	GET	System overview (total traces, success rate, active sessions)	❌
`/api/v1/monitor/traces`	GET	Recent trace list summary	❌
`/api/v1/monitor/traces/{trace_id}`	GET	Single trace details (including steps)	❌
`/api/v1/monitor/agents`	GET	Each Agent's current state	❌
`/api/v1/monitor/sessions`	GET	Active session list	❌

Operations Management¶

Endpoint	Method	Description	Auth
`/api/v1/operations/experiments`	POST	Create an experiment (overwrites if it exists)	❌
`/api/v1/operations/experiments`	GET	List all experiments	❌
`/api/v1/operations/experiments/{name}/results`	GET	Query experiment results	❌
`/api/v1/operations/experiments/{name}/metrics`	POST	Record one metric	❌
`/api/v1/operations/dashboard`	GET	Operations dashboard (30s cache)	❌
`/api/v1/operations/release-checklist`	GET	Go-live checklist report	❌

Escalation¶

Endpoint	Method	Description	Auth
`/api/v1/escalation/solution`	POST	Human-entered solution (pending review)	✅
`/api/v1/escalation/solutions/pending`	GET	List pending solutions	✅
`/api/v1/escalation/solutions/{solution_id}/approve`	POST	Approve and ingest as a FAQ	✅

Health Check¶

Endpoint	Method	Description	Auth
`/api/v1/health`	GET	Service health status (liveness probe)	❌

Endpoint Details¶

Chat Module¶

POST /api/v1/chat¶

Synchronous chat endpoint; returns a full ChatResponse after multi-agent orchestration.

Request body (ChatRequest):

Field	Type	Required	Description
`message`	string	✅	User message
`session_id`	string	❌	Session ID; auto-created when empty
`channel`	string	❌	Channel: web/app/wechat/dingtalk/api, default `web`
`user_id`	string	❌	User ID

Response body (ChatResponse):

{
  "session_id": "sess-abc123",
  "reply": "Hello, I have found the relevant content for you...",
  "status": "ok",
  "data": {
    "intent": "knowledge_qa",
    "sources": [{"source": "product_manual.md", "score": 0.92}],
    "escalate_to_human": false,
    "escalation_card": null,
    "turn_count": 1,
    "failed_attempts": 0,
    "emotion_score": 0.8,
    "sub_tasks": []
  }
}

curl example

curl -X POST http://localhost:8000/api/v1/chat \
  -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "message": "What is the return process?",
    "session_id": "sess-abc123",
    "channel": "web"
  }'

POST /api/v1/chat/stream¶

SSE streaming chat; event sequence meta → token (multiple) → done.

Event format: event: <type>\ndata: <json>\n\n

Event Type	data Fields	Description
`meta`	`intent`, `sources`, `escalate?`	Metadata; the first event
`token`	`content`	Streaming token chunk
`done`	`turn_count`, `escalate`, `answer`	Full answer and escalation flag
`error`	`message`	Error event (HTTP still 200)

curl example

curl -N -X POST http://localhost:8000/api/v1/chat/stream \
  -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{"message": "What is the return process?"}'

Nginx buffering configuration

The streaming endpoint already sets X-Accel-Buffering: no, but if buffering still occurs behind Nginx/CDN, disable it explicitly in the Nginx config:

location /api/v1/chat/stream {
    proxy_buffering off;
    proxy_cache off;
    chunked_transfer_encoding on;
}

Agent Assist¶

GET /api/v1/agent/sessions/pending¶

Returns all pending sessions, sorted by EscalationPriority descending.

Response body: List[AgentSessionSummary]

[
  {
    "session_id": "sess-abc",
    "user_id": "u-001",
    "channel": "web",
    "agent_status": "pending",
    "priority": "urgent",
    "turn_count": 3,
    "created_at": "2026-07-03T08:00:00+00:00"
  }
]

GET /api/v1/agent/sessions/{session_id}¶

Returns session details, including the EscalationCard and full history. Rebuilt on the fly on cache miss.

Path parameters:

Parameter	Type	Description
`session_id`	string	Session ID

Status codes:

Status Code	Description
200	Success
404	Session does not exist

POST /api/v1/agent/sessions/{session_id}/accept¶

Agent accepts the session. CAS ensures only one concurrent acceptance succeeds.

Request body (AcceptRequest, optional):

Field	Type	Description
`agent_id`	string	Agent ID

Status codes:

Status Code	Description
200	Acceptance succeeded
404	Session does not exist
409	Session is already `assigned`/`resolved`; cannot accept

POST /api/v1/agent/sessions/{session_id}/messages¶

Agent sends a message, appended to history.

Request body (AgentMessageRequest):

Field	Type	Required	Description
`content`	string	✅	Message content

Response body (AgentMessageResponse):

{
  "message_id": "msg-uuid",
  "timestamp": "2026-07-03T08:30:00+00:00",
  "role": "assistant"
}

Status codes:

Status Code	Description
200	Sent successfully
404	Session does not exist
409	Current state is not `assigned`; cannot send

Knowledge recommendation assist; reuses HybridRetriever.retrieve.

Request body (KnowledgeRecommendRequest):

Field	Type	Required	Default	Description
`query`	string	✅	—	Query text
`top_k`	int	❌	5	Number of chunks to return

Response body (KnowledgeRecommendResponse):

{
  "chunks": [
    {"content": "Return process...", "score": 0.92, "source": "return_policy.md"}
  ],
  "total": 1
}

Retrieval failure fallback

On retrieval exception, it degrades to an empty chunks list without error, ensuring the agent workbench is not interrupted.

POST /api/v1/agent/sessions/{session_id}/business-assist¶

Business query assist; reuses BusinessAgent.execute (with masking).

Request body (BusinessAssistRequest):

Field	Type	Required	Description
`query`	string	✅	Business query text

Response body (BusinessAssistResponse):

{
  "result": {
    "reply": "Your order #12345 has shipped...",
    "data": {"order_id": "12345", "phone_masked": "138****8888"},
    "error": null,
    "need_confirmation": false,
    "scene": "order_query"
  },
  "masked_fields": ["phone_masked"]
}

POST /api/v1/agent/sessions/{session_id}/resolve¶

Marks the session as resolved. CAS guards assigned → resolved.

Request body (ResolveRequest, optional):

Field	Type	Description
`note`	string	Resolution note

Status codes:

Status Code	Description
200	Marked successfully
404	Session does not exist
409	Current state is not `assigned`

POST /api/v1/agent/sessions/{session_id}/solution¶

Records a human solution, consolidated as a FAQ candidate (enters the pending review queue).

Request body (SolutionSubmitRequest):

Field	Type	Required	Description
`question`	string	✅	User question (`min_length=1`)
`solution`	string	✅	Solution (`min_length=1`)
`intent`	string	❌	Intent label; recognized by the system when empty

Response body (HumanSolutionRecord):

{
  "solution_id": "sol-uuid",
  "session_id": "sess-abc",
  "question": "How do I return a product?",
  "solution": "Please click Return on the order page...",
  "intent": "knowledge_qa",
  "status": "pending",
  "created_at": "2026-07-03T08:30:00+00:00"
}

Status codes:

Status Code	Description
200	Recorded successfully
404	Session does not exist
422	Empty `question`/`solution` triggers validation failure

Knowledge Base Management¶

POST /api/v1/knowledge/ingest¶

Upload a document for ingestion (multipart).

Request body (form-data):

Field	Type	Required	Default	Description
`file`	file	✅	—	File to ingest
`product_category`	string	❌	`unknown`	Product category
`applicable_version`	string	❌	`latest`	Applicable version
`knowledge_type`	string	❌	—	faq/policy/doc/tutorial/ticket
`published_at`	string	❌	—	Publish time ISO8601
`register`	bool	❌	false	Whether to register to the document registry
`validate_quality`	bool	❌	false	Whether to run quality checks at ingestion

Response body (IngestResult):

{
  "doc_id": "doc-uuid",
  "source": "return_policy.md",
  "chunk_count": 12,
  "status": "success",
  "errors": []
}

curl example

curl -X POST http://localhost:8000/api/v1/knowledge/ingest \
  -H "X-API-Key: $API_KEY" \
  -F "file=@return_policy.md" \
  -F "product_category=electronics" \
  -F "knowledge_type=policy" \
  -F "register=true"

GET /api/v1/knowledge/stats¶

Returns knowledge base statistics.

Response body (KnowledgeStats):

{
  "total_chunks": 1280,
  "total_documents": 15,
  "collection_name": "knowledge_base",
  "embedding_dim": 1024
}

GET /api/v1/knowledge/documents¶

Paginated list of registered documents.

Query parameters:

Parameter	Type	Default	Range	Description
`limit`	int	20	1–200	Page size
`offset`	int	0	≥0	Starting offset

GET /api/v1/knowledge/documents/{doc_id}¶

Query a single document's details, including full version history.

Response body (DocumentDetail):

{
  "doc_id": "doc-uuid",
  "source": "return_policy.md",
  "current_version": "v2",
  "status": "active",
  "created_at": "2026-06-01T00:00:00+00:00",
  "updated_at": "2026-07-01T00:00:00+00:00",
  "versions": [
    {"version": "v1", "doc_hash": "abc", "status": "superseded", "chunk_count": 10, "created_at": "..."},
    {"version": "v2", "doc_hash": "def", "status": "active", "chunk_count": 12, "created_at": "..."}
  ]
}

Document not found

When the document does not exist, an empty detail with status: "not_found" is returned; HTTP remains 200, and the frontend prompts accordingly.

DELETE /api/v1/knowledge/documents/{doc_id}¶

Delete a document by doc_id; removes all chunks of the document from the vector store.

Response body (DeleteResult):

{"doc_id": "doc-uuid", "deleted_chunks": 12, "success": true}

POST /api/v1/knowledge/quality/check¶

Run a batch quality inspection on ingested content.

Request body (QualityCheckRequest):

Field	Type	Description
`source`	string	Filter by source (optional)
`doc_id`	string	Filter by doc_id (optional)

Response body (QualityReport): includes inspection results such as duplicate chunks, terminology hit rate, and sensitive word hits.

POST /api/v1/knowledge/documents/{doc_id}/rollback¶

Roll back a document to a specified version.

Request body (RollbackRequest):

Field	Type	Description
`target_version`	string	Target version number

POST /api/v1/knowledge/canary/ingest & /canary/compare¶

Two canary verification endpoints; both use CanaryRequest as the request body:

Field	Type	Description
`doc_id`	string	Document ID
`version`	string	Target version
`sample_queries`	list[string]	Sample queries for comparison

Document Updates¶

POST /api/v1/update/full¶

Trigger a full update (monthly).

Request body (UpdateRequest):

Field	Type	Description
`dir_path`	string	Directory to scan
`extensions`	list[string]	File extension filter

POST /api/v1/update/incremental¶

Trigger an incremental update (weekly); processes only new files or files whose hash changed.

POST /api/v1/update/file¶

Single-file real-time update.

Request body (UpdateSingleFileRequest):

Field	Type	Description
`file_path`	string	Absolute file path
`metadata`	dict	Metadata overrides

GET /api/v1/update/status¶

Returns the most recent update result; empty if never run.

Retrieval Evaluation¶

POST /api/v1/evaluation/run¶

Trigger an evaluation run.

Request body (EvaluationRunRequest):

Field	Type	Default	Description
`testset_path`	string	Built-in 30 cases	External test set path
`top_k`	int	`rerank_top_k`	Retrieval top_k

Response body (EvaluationReport): includes report_id, total_queries, recall@5, hit_rate, and other metrics.

GET /api/v1/evaluation/reports¶

List historical report summaries, sorted by time descending.

GET /api/v1/evaluation/reports/{report_id}¶

Query a single report's details; returns 404 if it does not exist.

Performance Monitoring¶

GET /api/v1/performance/metrics¶

Returns comprehensive performance metrics.

Response body (MetricsResponse):

{
  "metrics": {
    "cache_hit_rate": 0.65,
    "concurrent_requests": 3,
    "model_router_stats": {"small_llm": 120, "main_llm": 30},
    "avg_response_ms": 1850
  }
}

GET /api/v1/performance/cache/stats¶

Returns hot cache statistics: hit/miss counts, hit rate, current entry count, LRU evictions.

POST /api/v1/performance/cache/invalidate¶

Clear the hot cache (call after knowledge base updates).

Response body (InvalidateResult):

{"success": true, "cleared": 42, "message": "Cleared 42 cache entries"}

Must call after knowledge base updates

After ingesting/deleting/rolling back knowledge base entries, you must call this endpoint to clear the cache; otherwise HotQueryCache will return stale replies.

Observability¶

GET /api/v1/observability/circuit-breakers¶

Returns a dict of name → CircuitBreakerStats, including state, failure count, and last failure time.

POST /api/v1/observability/circuit-breakers/{name}/reset¶

Manually reset a specified circuit breaker to CLOSED. Returns 404 if the breaker does not exist.

GET /api/v1/observability/alerts¶

Query the alert list, supports level / source / since filters.

Query parameters:

Parameter	Type	Description
`level`	string	info/warn/error/critical
`source`	string	Source, e.g. `token_usage`
`since`	string	ISO8601 start time

GET /api/v1/observability/health¶

Run all health checks and return an aggregate report.

GET /api/v1/observability/token-usage¶

Returns Token usage statistics for the specified window.

Query parameters:

Parameter	Type	Default	Values
`window`	string	`hour`	minute/hour/day

Monitoring (Agent dimension)¶

GET /api/v1/monitor/overview¶

Returns a system overview: total traces, success rate, average duration, active sessions.

GET /api/v1/monitor/traces¶

Returns recent trace list summary (without steps).

Query parameters:

Parameter	Type	Default	Range
`limit`	int	50	1–200

GET /api/v1/monitor/traces/{trace_id}¶

Returns a single trace's details (including steps and sub_tasks). Returns 404 if not found.

GET /api/v1/monitor/agents¶

Returns each Agent's current state (including uncalled agents with a call count of 0).

GET /api/v1/monitor/sessions¶

Returns the active session list, sorted by last-active time descending.

Operations Management¶

POST /api/v1/operations/experiments¶

Create an experiment; if it already exists, it is overwritten and historical metrics are cleared. Returns 201.

Request body (CreateExperimentRequest): includes experiment name, variant definitions, and traffic split.

GET /api/v1/operations/experiments¶

List all experiments.

GET /api/v1/operations/experiments/{name}/results¶

Query experiment results; returns 404 if the experiment does not exist.

POST /api/v1/operations/experiments/{name}/metrics¶

Record one experiment metric; returns 204. Recording is allowed even if the experiment does not exist, for replay.

GET /api/v1/operations/dashboard¶

Returns aggregated dashboard data, cached for 30 seconds. force_refresh=true bypasses the cache.

GET /api/v1/operations/release-checklist¶

Runs the go-live checklist and returns a report. Each check runs independently; a failure does not interrupt the others.

Escalation¶

POST /api/v1/escalation/solution¶

A human agent enters a solution (pending review).

Request body (HumanSolutionRequest):

Field	Type	Required	Description
`session_id`	string	✅	Session ID
`question`	string	✅	User question
`solution`	string	✅	Solution
`intent`	string	❌	Intent label

GET /api/v1/escalation/solutions/pending¶

List all pending solutions.

POST /api/v1/escalation/solutions/{solution_id}/approve¶

Approve a solution and ingest it as a FAQ. Returns 404 if the solution does not exist or ingestion fails.

Closed-loop value

After approval, the next time the bot retrieves a similar question, it can match this solution, forming the closed loop of "human handles → consolidate → bot answers next time".

Health Check¶

GET /api/v1/health¶

Liveness probe endpoint; not authenticated.

Response body (HealthResponse):

{
  "status": "ok",
  "app": "Intelligent Customer Service System",
  "version": "0.1.0",
  "timestamp": "2026-07-03T08:30:00+00:00"
}

Appendix¶

Endpoint Authentication Summary¶

flowchart LR
    A[Request] --> B{API_KEY non-empty?}
    B -- No --> C[Development mode: no auth]
    B -- Yes --> D{Endpoint requires auth?}
    D -- No --> E[Pass through directly<br/>Performance/Observability/Monitor/Operations/Health]
    D -- Yes --> F{X-API-Key matches?}
    F -- Yes --> E
    F -- No --> G[401 Unauthorized]

OpenAPI Documentation¶

FastAPI auto-generated interactive documentation:

Swagger UI: http://localhost:8000/docs
ReDoc: http://localhost:8000/redoc

API Reference¶

Authentication¶

Unified Error Response¶

Endpoint List Overview¶

Chat Module¶

Agent Assist¶

Knowledge Base Management¶

Document Updates¶

Ticket Mining and Retrieval Tuning¶

Retrieval Evaluation¶

Performance Monitoring¶

Observability¶

Monitoring (Agent dimension)¶

Operations Management¶

Escalation¶

Health Check¶

Endpoint Details¶

Chat Module¶

POST /api/v1/chat¶

POST /api/v1/chat/stream¶

Agent Assist¶

GET /api/v1/agent/sessions/pending¶

GET /api/v1/agent/sessions/{session_id}¶

POST /api/v1/agent/sessions/{session_id}/accept¶

POST /api/v1/agent/sessions/{session_id}/messages¶

POST /api/v1/agent/sessions/{session_id}/knowledge-recommend¶

POST /api/v1/agent/sessions/{session_id}/business-assist¶

POST /api/v1/agent/sessions/{session_id}/resolve¶

POST /api/v1/agent/sessions/{session_id}/solution¶

Knowledge Base Management¶

POST /api/v1/knowledge/ingest¶

GET /api/v1/knowledge/stats¶

GET /api/v1/knowledge/documents¶

GET /api/v1/knowledge/documents/{doc_id}¶

DELETE /api/v1/knowledge/documents/{doc_id}¶

POST /api/v1/knowledge/quality/check¶

POST /api/v1/knowledge/documents/{doc_id}/rollback¶

POST /api/v1/knowledge/canary/ingest & /canary/compare¶

Document Updates¶

POST /api/v1/update/full¶

POST /api/v1/update/incremental¶

POST /api/v1/update/file¶

GET /api/v1/update/status¶

Retrieval Evaluation¶

POST /api/v1/evaluation/run¶

GET /api/v1/evaluation/reports¶

GET /api/v1/evaluation/reports/{report_id}¶

Performance Monitoring¶

GET /api/v1/performance/metrics¶

GET /api/v1/performance/cache/stats¶

POST /api/v1/performance/cache/invalidate¶

Observability¶

GET /api/v1/observability/circuit-breakers¶

POST /api/v1/observability/circuit-breakers/{name}/reset¶

GET /api/v1/observability/alerts¶

GET /api/v1/observability/health¶

GET /api/v1/observability/token-usage¶

Monitoring (Agent dimension)¶

GET /api/v1/monitor/overview¶

GET /api/v1/monitor/traces¶

GET /api/v1/monitor/traces/{trace_id}¶

GET /api/v1/monitor/agents¶

GET /api/v1/monitor/sessions¶

Operations Management¶

POST /api/v1/operations/experiments¶

GET /api/v1/operations/experiments¶

GET /api/v1/operations/experiments/{name}/results¶

POST /api/v1/operations/experiments/{name}/metrics¶

GET /api/v1/operations/dashboard¶

GET /api/v1/operations/release-checklist¶

Escalation¶

POST /api/v1/escalation/solution¶

GET /api/v1/escalation/solutions/pending¶

POST /api/v1/escalation/solutions/{solution_id}/approve¶

Health Check¶

GET /api/v1/health¶

Appendix¶

Endpoint Authentication Summary¶

OpenAPI Documentation¶

Related Documentation¶