Query API

LynxDB exposes three query-facing paths:

• POST /api/v1/query for JSON-envelope query execution
• POST /api/v1/query/stream for NDJSON export
• GET /api/v1/query/explain for parse and plan inspection

Async work created by POST /query is tracked under /api/v1/query/jobs/....

POST /query

Execute a LynxFlow query and return either final results or a job handle.

Request Body

Field	Type	Required	Description
q	string	Yes*	Query text
query	string	Yes*	Alias for q
from	string	No	Start time bound
earliest	string	No	Alias for from
to	string	No	End time bound
latest	string	No	Alias for to
limit	integer	No	Result row limit. Defaults to query.default_result_limit, capped by query.max_result_limit
offset	integer	No	Offset for paginated event or aggregate results
wait	number	No	Execution mode selector
profile	string	No	Profiling level: basic, full, or trace
variables	object	No	Replaces $name tokens in the query with quoted, escaped string values
format	string	No	Response format. Only empty or json are accepted on this endpoint. Use POST /api/v1/query/stream for NDJSON output
language	string	No	Query language. Only lynxflow (or omitted) is accepted

* One of q or query is required.

Query Language

All queries execute as LynxFlow. The language field is optional; when present, only "lynxflow" is accepted. Requesting "spl2" returns 400 VALIDATION_ERROR with migration guidance:

{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "language \"spl2\" is no longer supported; migrate queries to LynxFlow — see CHANGELOG for migration guide"
  }
}

Successful responses include meta.language: "lynxflow".

Execution Modes

wait controls when the server returns:

wait value	Mode	Behavior
omitted or null	sync window	Wait until completion. If the query does not finish within query.sync_timeout (default 30s), LynxDB detaches the job and returns 202 Accepted with a job handle
0	async	Return 202 Accepted immediately with a job handle
positive number	hybrid	Wait up to N seconds. If the query is still running, return the same 202 Accepted job handle

Event Result Example

curl -s localhost:3100/api/v1/query \
  -H 'Content-Type: application/json' \
  -d '{
    "q": "from main | where level == \"error\" | head 2",
    "from": "-1h"
  }' | jq .

{
  "data": {
    "type": "events",
    "events": [
      {
        "_time": "2026-02-14T14:52:01.234Z",
        "_raw": "level=ERROR status=502 uri=/api/users",
        "source": "nginx",
        "host": "web-01",
        "level": "ERROR",
        "status": 502,
        "uri": "/api/users"
      }
    ],
    "total": 1247,
    "has_more": true
  },
  "meta": {
    "took_ms": 89,
    "scanned": 12400000,
    "query_id": "qry_7f3a...",
    "language": "lynxflow"
  }
}

Aggregate Result Example

curl -s localhost:3100/api/v1/query \
  -H 'Content-Type: application/json' \
  -d '{
    "q": "from main | where source == \"nginx\" and status >= 500 | stats count() by uri | sort -count | head 10",
    "from": "-1h"
  }' | jq .

{
  "data": {
    "type": "aggregate",
    "columns": ["uri", "count"],
    "rows": [
      ["/api/v1/users", 1247],
      ["/api/v1/orders", 893]
    ],
    "total_rows": 42,
    "has_more": true
  },
  "meta": {
    "took_ms": 34,
    "scanned": 12400000,
    "query_id": "qry_8b2c..."
  }
}

Async or Promoted Job Handle Example

When a query is explicitly async, or when it misses the sync window, POST /query returns 202 Accepted:

{
  "data": {
    "type": "job",
    "job_id": "qry_9c1d4e",
    "status": "running"
  },
  "meta": {
    "query_id": "qry_9c1d4e"
  }
}

Current handlers return a minimal job handle only. They do not include query, from, to, or partial result rows in the 202 response body.

Variable Substitution

variables replaces $name tokens with quoted string literals before planning:

curl -s localhost:3100/api/v1/query \
  -H 'Content-Type: application/json' \
  -d '{
    "q": "from main | where source == $source and level == $level | head 5",
    "variables": {
      "source": "nginx",
      "level": "error"
    }
  }' | jq .

Response Behavior

• POST /query always returns the standard JSON envelope.
• For NDJSON export, use POST /query/stream.
• format values other than json are rejected with 400 VALIDATION_ERROR.
• Removed SPL2 stages (eval, rex, table, timechart, ...) are rejected at parse time with 400 INVALID_QUERY and a suggestion for the LynxFlow equivalent.

Error Responses

Status	Code	When
400	INVALID_JSON	Request body is not valid JSON
400	INVALID_QUERY	Parse or planning failed
400	QUERY_TOO_LARGE	Query exceeds query.max_query_length
400	QUERY_MEMORY_EXCEEDED	Query exceeded its per-query memory budget
400	VALIDATION_ERROR	Query is missing, format is not json, or language is not lynxflow
401	AUTH_REQUIRED / INVALID_TOKEN	Authentication failure
403	FORBIDDEN	Token lacks query scope
429	TOO_MANY_REQUESTS	Query concurrency limit reached
503	QUERY_POOL_EXHAUSTED	Global query memory pool is exhausted

GET /query

GET convenience variant for simple queries.

Query Parameters

Parameter	Required	Description
q	Yes	Query text
from	No	Start time bound
to	No	End time bound
limit	No	Result row limit
format	No	Optional response format. Only empty or json are accepted
language	No	Query language. Only lynxflow (or omitted) is accepted

curl -s "localhost:3100/api/v1/query?q=from+main+%7C+head+5&limit=5" | jq .

GET /query returns the same JSON envelope and result shapes as POST /query. As with POST /query, format=csv, format=ndjson, and other non-JSON values are rejected.

POST /query/stream

Stream query results as newline-delimited JSON.

This path is for export and pipeline use, not job management.

Request Body

POST /query/stream accepts a subset of the POST /query request body:

• q or query
• from or earliest
• to or latest
• variables
• language (only lynxflow or omitted)

limit, offset, wait, profile, and format are not silently ignored on this path. If any of them are present, the handler returns 400 VALIDATION_ERROR.

Example

curl -s localhost:3100/api/v1/query/stream \
  -H 'Content-Type: application/json' \
  -d '{"q":"from main | where level == \"error\"","from":"-24h"}'

Example response:

{"_time":"2026-02-14T14:52:01Z","_raw":"level=ERROR msg=\"timeout\"","level":"ERROR","message":"timeout"}
{"_time":"2026-02-14T14:51:58Z","_raw":"level=ERROR msg=\"refused\"","level":"ERROR","message":"refused"}
{"__meta":{"total":8432,"scanned":12400000,"took_ms":342}}

Behavior

• The response content type is application/x-ndjson.
• There is no default result limit on this path.
• The final line is {"__meta": ...} with stream summary data.
• If a streaming error occurs after output has started, LynxDB writes a final {"__error": ...} line.
• Client disconnect cancels the streaming query.

GET /query/explain

Parse and explain a query without executing it.

Query Parameters

Parameter	Required	Description
q	Yes*	Query text
query	Yes*	Alias for q
from	No	Start time bound (accepted for compatibility; does not change the plan text)
to	No	End time bound (accepted for compatibility; does not change the plan text)
analyze	No	Accepted for compatibility; the LynxFlow explain response does not currently include execution stats
language	No	Query language. Only lynxflow (or omitted) is accepted

* One of q or query is required.

Valid Response Example

curl -s "localhost:3100/api/v1/query/explain?q=from%20main%20%22error%22%20%7C%20head%2010" | jq .

The response carries the optimized plan as text in lynxflow_plan, including desugar rewrites and the optimizer rules that fired:

{
  "data": {
    "is_valid": true,
    "lynxflow_plan": "1. Limit(10)\n  2. Filter(contains(_raw, \"error\"))\n    3. Scan(main)\n      bloom_term: \"error\"\n\nAnnotations:\n  Desugar rewrites:\n    search-sugar: \"error\" => where contains(_raw, \"error\")\n  Optimizer rules:\n    predicate-pushdown (1x)\n",
    "errors": []
  },
  "meta": {
    "language": "lynxflow"
  }
}

Invalid Response Example

GET /query/explain still returns 200 OK for invalid queries:

{
  "data": {
    "is_valid": false,
    "errors": [
      {
        "message": "lynxflow.Explain: parse: unknown stage \"staats\", did you mean \"stats\"?"
      }
    ]
  },
  "meta": {
    "language": "lynxflow"
  }
}

GET /query/jobs

List active and recently completed jobs.

curl -s localhost:3100/api/v1/query/jobs | jq .

{
  "data": {
    "jobs": [
      {
        "job_id": "qry_9c1d4e",
        "query": "from main | head 5",
        "status": "running",
        "created_at": "2026-02-14T14:50:00Z"
      }
    ]
  }
}

Notes

• GET /query/jobs accepts an optional status query parameter. Supported values are running, done, error, canceled, plus aliases complete, failed, and cancelled.
• Completed, errored, and canceled jobs are garbage-collected after query.job_ttl (default 5m).

GET /query/jobs/{id}

Fetch a specific job.

Running Job

{
  "data": {
    "type": "job",
    "job_id": "qry_9c1d4e",
    "status": "running",
    "query": "from main | head 5",
    "progress": {
      "phase": "scanning_segments",
      "segments_total": 30,
      "segments_scanned": 12,
      "segments_dispatched": 18,
      "segments_skipped_index": 4,
      "segments_skipped_time": 2,
      "segments_skipped_stats": 1,
      "segments_skipped_bloom": 3,
      "segments_skipped_range": 0,
      "buffered_events": 1200,
      "rows_read_so_far": 120000,
      "elapsed_ms": 18400
    }
  }
}

Completed Job

{
  "data": {
    "type": "job",
    "job_id": "qry_9c1d4e",
    "status": "done",
    "query": "from main | stats count() by source",
    "created_at": "2026-02-14T14:50:00Z",
    "completed_at": "2026-02-14T14:50:37Z",
    "results": {
      "type": "aggregate",
      "columns": ["source", "count"],
      "rows": [
        ["nginx", 142847]
      ],
      "total_rows": 5,
      "has_more": false
    }
  },
  "meta": {
    "took_ms": 37200,
    "scanned": 84700000000,
    "query_id": "qry_9c1d4e"
  }
}

Errored or Canceled Job

{
  "data": {
    "type": "job",
    "job_id": "qry_d4e1f2",
    "status": "error",
    "error": {
      "code": "QUERY_MEMORY_EXCEEDED",
      "message": "query exceeded memory budget"
    }
  }
}

Canceled jobs use the same envelope shape, with status: "canceled" and an error message such as canceled by user.

Error Responses

Status	Code	Description
404	NOT_FOUND	Job ID not found

DELETE /query/jobs/{id}

Cancel a job.

This endpoint requires admin scope when auth is enabled.

curl -X DELETE localhost:3100/api/v1/query/jobs/qry_9c1d4e | jq .

{
  "data": {
    "job_id": "qry_9c1d4e",
    "status": "canceled",
    "canceled": true,
    "completed_at": "2026-02-14T14:50:37Z"
  }
}

canceled reports whether this request actually stopped a running job. If the job had already finished, the endpoint still returns 200 OK, but canceled is false and status reflects the final state.

GET /query/jobs/{id}/stream

Stream job progress over Server-Sent Events (SSE).

curl -N localhost:3100/api/v1/query/jobs/qry_9c1d4e/stream

Event Types

Event	When	Data
progress	About once per second while the job is running	Progress information. If preview rows are available, they are embedded in the same event as preview and preview_version
complete	Job finished successfully	Final payload shaped as {data: ..., meta: ...}
failed	Job errored	{code, message}
canceled	Job was canceled	Progress snapshot

Example Stream

event: progress
data: {"phase":"scanning","segments_total":30,"scanned":12,"percent":40,"elapsed_ms":5000}

event: progress
data: {"phase":"scanning","segments_total":30,"scanned":21,"percent":70,"elapsed_ms":9000,"preview":[{"source":"nginx","count":142000}],"preview_version":1}

event: complete
data: {"data":{"type":"aggregate","columns":["source","count"],"rows":[["nginx",284000]],"total_rows":1,"has_more":false},"meta":{"took_ms":10400,"scanned":847000,"stats":{"rows_scanned":847000}}}

The current implementation does not emit a separate partial event type. Preview rows, when present, are attached to progress events.

Related

• REST API overview for authentication, envelopes, and endpoint map
• Live Tail & Histogram for SSE log streaming
• lynxdb query for the CLI behavior on top of these endpoints
• LynxFlow v2 Reference for LynxFlow syntax