ollama_multi_file_refactor_propose

Propose multi-file refactor. REFACTOR. Coordinated multi-file refactor PLAN — NO WRITES. Server reads each file in files (1-20 paths, path validation via SOURCE_PATH_NOT_FOUND), hands the bodies + change_description to the Workhorse tier, and returns {per_file_changes:[{file, before_summary, after_summary, risk_level, change_kinds[]}], cross_file_impact, affected_imports:[{from, to, files[]}], verification_steps, weak}. change_kinds are normalized from {rename, signature-change, import-update, move, delete, new}. Files the model invents (not in input) are stripped. Thin output → weak=true. Use BEFORE touching files so Claude can see a coordinated plan. Pair with ollama_refactor_plan for sequencing and ollama_batch_proof_check for verification.

Parameters

Parameter	Type	Required	Default	Description
files	any	yes	—	Source file paths the refactor touches. Server reads them — Claude does not preload. Min 1, max 20.
change_description	string	yes	—	What the refactor should accomplish, in 20-2000 chars. Be specific — vague descriptions produce vague plans (likely weak=true).
per_file_max_chars	integer	no	—	Chars to read per file (default 60_000).
backend	"cloud" \	"local"	no	—

Full input schema

The JSON Schema below is generated from the same zod schema the server validates against at the wire — it cannot drift from runtime behavior.

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "files": {
      "description": "Source file paths the refactor touches. Server reads them — Claude does not preload. Min 1, max 20."
    },
    "change_description": {
      "type": "string",
      "minLength": 20,
      "maxLength": 2000,
      "description": "What the refactor should accomplish, in 20-2000 chars. Be specific — vague descriptions produce vague plans (likely weak=true)."
    },
    "per_file_max_chars": {
      "description": "Chars to read per file (default 60_000).",
      "type": "integer",
      "minimum": 1000,
      "maximum": 200000
    },
    "backend": {
      "description": "Optional per-call backend directive (v2.9). 'cloud' escalates THIS call to Ollama Cloud — works in cloud standby (OLLAMA_API_KEY set, OLLAMA_CLOUD_PRIMARY unset) and cloud-primary modes; errors with CLOUD_NOT_CONFIGURED when no cloud is configured (never silently runs local while claiming escalation). 'local' pins the call to the local backend (zero egress) even under cloud-primary. Omit for the mode default. Escalated calls disclose egress loudly and carry backend provenance on the envelope. Synthesis quality on this tool scales sharply with model class — on a repo-scale job the local 8B is often the binding constraint, not the evidence. Tools where a local model is genuinely adequate (ollama_summarize_fast, ollama_classify, ollama_triage_logs, and every no-LLM tool) deliberately do NOT offer this field.",
      "type": "string",
      "enum": [
        "cloud",
        "local"
      ]
    }
  },
  "required": [
    "files",
    "change_description"
  ]
}

Envelope

Every tool returns the standard envelope — tier_used, model, tokens_in/tokens_out, elapsed_ms, backend provenance, warnings. See Envelope & tiers.

Source

Schema + handler: src/tools/multiFileRefactorPropose.ts · registration: src/index.ts

---

Back to the Tool Reference.