Skip to content
ollama-intern-mcp

ollama_extract

ollama_extract pulls structured fields out of free-form text. You provide the JSONSchema; the model fills it in. Output is constrained to round-trip through your schema, or the call fails loud with unparseable.

Tier — Workhorse

Section titled “Tier — Workhorse”

hermes3:8b on the default dev-rtx5080 profile, num_ctx: 8192. Larger context than Instant tier so multi-paragraph inputs fit.

When to use it

Section titled “When to use it”
  • Pulling fields out of commits / tickets / log lines / email threads
  • Converting prose specs into structured task lists
  • Normalizing messy CSV / JSON / log inputs into a consistent shape

When NOT to use it

Section titled “When NOT to use it”
  • The output is a single label → ollama_classify (faster, has confidence)
  • You need cross-file synthesis → ollama_research or ollama_corpus_answer
  • The input is a long file → use source_path mode (server reads it)

Schema

Section titled “Schema”
{
  // Exactly one of {text, source_path, items} must be provided.
  text?:        string;
  source_path?: string;
  items?: Array<{
    id: string;
    text: string;
  }>;


  schema:       Record<string, unknown>;  // JSONSchema the result must conform to
  hint?:        string;                    // optional field-by-field hint
  frame?:       string;                    // v2.2.0+ — what the extraction is FOR
  per_file_max_chars?: number;
  model?:       string;
}

Full source: src/tools/extract.ts.

Example call (single)

Section titled “Example call (single)”
{
  "tool": "ollama_extract",
  "arguments": {
    "text": "feat(corpus): surface :latest drift on refresh",
    "schema": {
      "type": "object",
      "properties": {
        "kind":  { "enum": ["feat", "fix", "chore", "docs", "refactor"] },
        "scope": { "type": "string" },
        "summary": { "type": "string" }
      },
      "required": ["kind", "summary"]
    }
  }
}

Returns:

{
  "result": {
    "extracted": {
      "kind": "feat",
      "scope": "corpus",
      "summary": "surface :latest drift on refresh"
    }
  },
  "tier_used": "workhorse",
  "...": "..."
}

Example call (batch)

Section titled “Example call (batch)”
{
  "tool": "ollama_extract",
  "arguments": {
    "schema": {
      "type": "object",
      "properties": {
        "severity": { "enum": ["low", "medium", "high"] },
        "service": { "type": "string" }
      }
    },
    "items": [
      { "id": "log-1", "text": "[ERROR] payments-svc: timeout after 30s" },
      { "id": "log-2", "text": "[WARN] auth-svc: token refresh slow (1.2s)" }
    ]
  }
}

Returns one envelope with per-item {id, ok, result|error} entries.

Example call (source_path mode)

Section titled “Example call (source_path mode)”

For a long file you don’t want Claude to pre-read:

{
  "tool": "ollama_extract",
  "arguments": {
    "source_path": "/repo/src/index.ts",
    "schema": { "type": "object", "properties": { "exported_symbols": { "type": "array" } } },
    "per_file_max_chars": 60000
  }
}

The server reads the file, truncates to per_file_max_chars, and runs extraction — Claude never sees the raw file content.

Common pitfalls

Section titled “Common pitfalls”

{ error: "unparseable" } returned in result. The model produced output that couldn’t round-trip through your JSONSchema. Most common causes: schema is too restrictive (missing additionalProperties: false that the model can satisfy), required field has no signal in the input, or the input is genuinely off-topic. Try simplifying the schema or adding a hint.

PATH_NOT_ALLOWED on source_path. The path is outside INTERN_ALLOWED_ROOTS. Either fix the path or extend the allow-list.

Schema too large. Massive nested schemas eat the context window. Split into multiple targeted extractions if you need to fill 30+ fields.

Two input modes provided. Returns SCHEMA_INVALID. Pick exactly one of text / source_path / items.

Section titled “Related tools”