Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Excurs1ons/MonoRelay/llms.txt

Use this file to discover all available pages before exploring further.

Every request proxied through MonoRelay is written to a local SQLite database with full token counts, latency metrics, and optional request/response body previews. The logs API lets you page through recent requests, retrieve a specific log entry with its full body, stream new entries in real time via SSE, filter by provider or model, and clear the log store when needed.

​
Authentication

All log endpoints require a valid JWT token: 
Authorization: Bearer <jwt>

​
Endpoints

​
GET /api/logs

Return a paginated list of recent log entries, newest first. Heavy fields (request_full, response_full) are stripped from the list view; fetch them individually via /api/logs/{id}. Query parameters:
ParameterTypeDefaultDescription
pageinteger1Page number (1-indexed)
page_sizeinteger20Items per page
limitinteger50Maximum total entries to load before pagination
curl "http://localhost:8787/api/logs?page=1&page_size=20" \
  -H "Authorization: Bearer <jwt>"
Response envelope: 
{
  "success": true,
  "data": [ /* array of log records */ ],
  "metadata": {
    "page": 1,
    "page_size": 20,
    "total": 143,
    "timestamp": "2026-05-17T10:00:00"
  }
}

​
GET /api/logs/

Retrieve a single log entry by ID, including the full request and response bodies. Log IDs are integers. Pending (in-flight streaming) requests use negative temporary IDs until finalized. 
curl http://localhost:8787/api/logs/42 \
  -H "Authorization: Bearer <jwt>"
Returns 404 if the ID is not found in the database or in the in-memory pending store.

​
GET /api/logs/stream

An SSE (Server-Sent Events) endpoint that pushes new and updated log entries in real time. The dashboard uses this to display live request activity without polling. 
curl -N http://localhost:8787/api/logs/stream \
  -H "Authorization: Bearer <jwt>"
The stream emits two event types:
Event typeWhen emitted
log_newA new request starts (immediately, before completion)
log_updateA pending entry is updated (first token received, stream finalized)
A heartbeat comment (: heartbeat) is sent every 20 seconds to keep the connection alive through proxies. Example SSE output: 
event: log_new
data: {"id": -1, "model": "claude-opus-4-5", "provider": "anthropic", "timestamp": 1747468800.0}

event: log_update
data: {"id": -1, "_real_id": 99, "status_code": 200, "latency_ms": 1342.5, "input_tokens": 84, "output_tokens": 210}

​
GET /api/logs/filtered

Return logs filtered by provider, model, status, or date range. Applies in-memory filtering after loading a larger window of recent entries. Query parameters:
ParameterTypeDescription
providerstringExact provider name match (e.g. anthropic)
modelstringCase-insensitive substring match on model name
statusstring"success" (status < 400) or "error" (status ≥ 400)
date_fromstringISO date string, e.g. 2026-05-01
date_tostringISO date string, e.g. 2026-05-17
limitintegerMax entries to scan (default 100)
pageintegerPage number
page_sizeintegerItems per page (default 20)
curl "http://localhost:8787/api/logs/filtered?provider=anthropic&status=error&date_from=2026-05-01" \
  -H "Authorization: Bearer <jwt>"

​
POST /api/logs/clear

Delete all log entries from the database and clear any in-memory pending entries. 
curl -X POST http://localhost:8787/api/logs/clear \
  -H "Authorization: Bearer <jwt>"

{"success": true, "message": "请求日志已清空"}

​
Log record fields

Each log entry returned by the API contains these fields:
​
id
integer
Database row ID. Negative values indicate pending (in-flight) entries not yet written to disk.
​
timestamp
number
Unix timestamp (float) of when the request was received.
​
model
string
The resolved model name as used for the upstream request.
​
provider
string
The provider that handled the request.
​
key_label
string
Label of the API key used, as configured in providers.<name>.keys[].label.
​
status_code
integer
HTTP status code returned by the upstream provider (e.g. 200, 429, 500).
​
latency_ms
number
Total request latency in milliseconds from receipt to final byte.
​
first_token_ms
number
Time to first token in milliseconds. Populated for streaming requests.
​
input_tokens
integer
Number of prompt tokens consumed.
​
output_tokens
integer
Number of completion tokens generated.
​
estimated_cost
number
Estimated USD cost calculated from configured cost_per_m_input / cost_per_m_output rates.
​
cache_hit_tokens
integer
Tokens served from the upstream prompt cache (Anthropic cache read).
​
cache_miss_tokens
integer
Tokens written to the upstream prompt cache (Anthropic cache creation).
​
streaming
boolean
true if the request used streaming mode.
​
error_message
string
Error description if the request failed, otherwise null.
​
error_type
string
Error category string, e.g. "rate_limit_error", "proxy_error".
​
request_preview
string
Truncated preview of the request body (length controlled by logging.content_preview_length). Stripped from list endpoints; available in detail view.
​
response_preview
string
Truncated preview of the response content. Updated in real time for streaming requests.
​
request_full
string
Full JSON request body. Only included in single-entry detail responses (GET /api/logs/{id}).
​
response_full
string
Full JSON response body. Only included in single-entry detail responses.
​
client_ip
string
Originating client IP address, resolved from x-real-ip, x-forwarded-for, or the direct connection.
​
user_agent
string
The User-Agent header sent by the calling client.

​
Logging configuration reference

Logging is configured in config.yml under the logging key:
FieldTypeDefaultDescription
enabledbooleantrueEnable or disable request logging entirely
db_pathstring./data/requests.dbPath to the SQLite database file
max_age_daysinteger30Automatically delete log entries older than this many days
content_preview_lengthinteger200Maximum characters to store in request/response preview fields
Example: 
logging:
  enabled: true
  db_path: ./data/requests.db
  max_age_days: 30
  content_preview_length: 200

Build docs developers (and LLMs) love

Get started for freeTalk to us