Index a workspace for semantic code search and query it — /rag/index, /rag/query, /rag/status.

MisarCoder's retrieval layer indexes a workspace's files into chunks and searches them by meaning, by keyword, or by both.

Base URL: https://api.misar.io/coder. All three endpoints authenticate with Authorization: Bearer <API_KEY>.

Workspaces are derived, not created

There is no "create workspace" call. A workspace id is derived from your API token plus the workspace_path you send, so the same path under a different token is a different workspace. Send the same workspace_path on every call for a given project.

Index files

POST/rag/index

Chunks and embeds a batch of files.

Request body

workspace_pathstringbodyrequired

Absolute path of the workspace on the client machine. Used to derive the workspace id.

filesArray<{path, content}>bodyrequired

Files to index. At most 200 per request.

workspace_namestringbody

Human-readable workspace name.

backgroundbooleanbodydefault: false

Queue the work instead of blocking. The response is then just an acknowledgement.

Response fields

statusstring

ok for a synchronous run, queued when background was true.

workspace_idstring

The derived workspace id.

chunksnumber

Chunks produced. Synchronous runs only.

embeddednumber

Chunks embedded. Synchronous runs only.

filesnumber

Files processed. Synchronous runs only.

Request

curl -X POST https://api.misar.io/coder/rag/index \
  -H "Authorization: Bearer $MISARCODER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workspace_path": "/Users/jane/acme-api",
    "workspace_name": "acme-api",
    "files": [{ "path": "src/server.ts", "content": "export const start = …" }]
  }'

200 — Indexed

{
  "status": "ok",
  "workspace_id": "ws_4c1a…",
  "chunks": 128,
  "embedded": 128,
  "files": 1
}

200 — Queued

{ "status": "queued", "workspace_id": "ws_4c1a…" }

Limits

Limit	Value
Files per request	200
File content size	512 KB each
Chunks per workspace	10,000

Status codes

Code	Meaning
`200`	Indexed or queued
`400`	Too many files, or one or more files exceed the size cap. The message names up to five offending paths.
`401`	`Missing authorization` or `Invalid API key`
`429`	Workspace chunk limit reached. Delete old files or contact support.
`500`	`Indexing failed`
`503`	RAG is unavailable — the server is not connected to a database

Query

POST/rag/query

Searches an indexed workspace.

Request body

workspace_pathstringbodyrequired

Same path you indexed under.

querystringbodyrequired

Natural-language or code query.

knumberbodydefault: 10

Results to return, between 1 and 50.

modestringbodydefault: hybrid

hybrid, vector, or lexical. An unrecognised value silently falls back to hybrid.

Response fields

statusstring

ok.

workspace_idstring

The derived workspace id.

modestring

The mode that actually executed — not necessarily the one you asked for.

query_rewrittenstring

The query after rewriting.

query_typestring

Classified query type. Hybrid mode only.

resultsobject[]

Matches with file_path, symbol, content, chunk_level, start_byte, plus rank for lexical results or distance for vector results.

statsobject

{ total_chunks, embedded_chunks, total_files } for the workspace.

Request

curl -X POST https://api.misar.io/coder/rag/query \
  -H "Authorization: Bearer $MISARCODER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workspace_path": "/Users/jane/acme-api",
    "query": "where is rate limiting applied",
    "k": 10,
    "mode": "hybrid"
  }'

200 — OK

{
  "status": "ok",
  "workspace_id": "ws_4c1a…",
  "mode": "hybrid",
  "query_rewritten": "rate limit middleware",
  "query_type": "lookup",
  "results": [
    {
      "file_path": "src/middleware/rate-limit.ts",
      "symbol": "applyRateLimit",
      "content": "export function applyRateLimit(…)",
      "chunk_level": "function",
      "start_byte": 412,
      "distance": 0.18
    }
  ],
  "stats": { "total_chunks": 128, "embedded_chunks": 128, "total_files": 1 }
}

`mode` in the response is what ran, not what you asked for

Requesting hybrid or vector against a workspace with no embedded chunks executes a lexical search and returns "mode": "lexical". Read the field back rather than assuming your requested mode was honoured.

Queries time out after 10 seconds. The query text is capped at 2000 characters.

Status codes

Code	Meaning
`200`	Results returned
`401`	`Missing authorization` or `Invalid API key`
`422`	`k` outside 1–50, or a missing required field
`500`	`Query failed`
`504`	`Query timed out`

Status

POST/rag/status

Index statistics for a workspace. Use it to decide whether a re-index is needed before querying.

Request body

workspace_pathstringbodyrequired

Absolute path of the workspace on the client machine.

Response fields

statusstring

ok.

workspace_idstring

The derived workspace id.

total_chunksnumber

Chunks stored. Compare against the 10,000 ceiling.

embedded_chunksnumber

Chunks with embeddings. Below total_chunks means vector search is incomplete.

total_filesnumber

Files represented in the index.

Request

curl -X POST https://api.misar.io/coder/rag/status \
  -H "Authorization: Bearer $MISARCODER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "workspace_path": "/Users/jane/acme-api" }'

200 — OK

{
  "status": "ok",
  "workspace_id": "ws_4c1a…",
  "total_chunks": 128,
  "embedded_chunks": 128,
  "total_files": 1
}

Status codes

Code	Meaning
`200`	Statistics returned
`401`	`Missing authorization` or `Invalid API key`
`503`	RAG is unavailable — the server is not connected to a database

RAG needs a database

All three endpoints return 503 when the deployment has no database configured. This is a deployment condition, not a transient fault — retrying will not help.

RAG

Index files

Limits

Status codes

Query

Status codes

Status

Status codes

On this page