Skip to main content

Documentation Index

Fetch the complete documentation index at: https://kode-f177b001.mintlify.app/llms.txt

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

kode verify runs a set of proposed code changes through Kode’s full verification pipeline without applying anything to disk. Give it a JSON file of hunks or file contents, and it will check each change for syntax correctness, valid imports, call graph consistency, blast radius limits, architectural rule compliance, and security vulnerabilities. Use it to validate patches generated outside of Kode, to build CI gates, or to inspect what kode generate produced before committing.

​
Synopsis

kode verify --input <file>

​
Input modes

kode verify accepts two JSON structures:

​
Hunk mode

Use hunk mode when you have structured hunks scoped to specific files — the format produced by kode generate. Kode applies the hunks in-memory, then runs the full verification pipeline on the resulting file contents. 
{
  "hunks": [
    {
      "id": "hunk-1",
      "file_path": "internal/user/handler.go",
      "action": "MODIFY",
      "target_symbol": "CreateUser",
      "anchor_text": "func CreateUser(w http.ResponseWriter, r *http.Request) {",
      "new_text": "func CreateUser(w http.ResponseWriter, r *http.Request) {\n\tif r.Body == nil { http.Error(w, \"empty body\", 400); return }",
      "explanation": "Add nil body guard to CreateUser"
    }
  ],
  "original_files": {
    "internal/user/handler.go": "<full original file contents>"
  }
}
The original_files map provides the pre-patch file contents so Kode can apply hunks accurately. Keys must match the file_path field in each hunk.

​
File mode

Use file mode when you have complete proposed file contents rather than diffs. Kode verifies the files directly without applying any patches. 
{
  "files": {
    "internal/user/handler.go": "<complete proposed file contents>",
    "internal/user/service.go": "<complete proposed file contents>"
  }
}
You can also embed block_architecture and architecture_rules in the JSON itself: 
{
  "files": { "...": "..." },
  "block_architecture": true,
  "architecture_rules": [
    {
      "forbidden_import_prefix": "internal/db",
      "allowed_in_packages": ["internal/repository"],
      "error_message": "Only repository packages may import internal/db"
    }
  ]
}

​
Flags

​
--input
string
required
Path to the JSON file containing hunks or proposed file contents. This flag is required.
kode verify --input patches.json
​
--project-dir
string
default:"cwd"
Project root directory. Kode resolves import paths and architecture rules relative to this directory.
​
--block-architecture
boolean
default:"false"
Treat architecture rule violations as hard failures. By default, architecture violations are reported but do not cause the pipeline to exit 1. Pass this flag to enforce them strictly.
kode verify --input patches.json --block-architecture
​
--log-dir
string
default:"<project-dir>/logs"
Directory where Kode appends the JSONL audit log entry for this run. The file is named kode.log. Kode creates the directory if it does not exist.
​
--model
string
Model identifier written to the audit log entry for telemetry purposes. Does not affect verification behaviour.

​
Exit codes

CodeMeaning
0All hunks or files passed every gate (PASS)
1One or more hunks failed at least one gate (FAIL), or the input file could not be read or parsed

​
Output

kode verify writes a verdict JSON object to stdout: 
{
  "task_id": "verify",
  "status": "PASS",
  "applied_hunks": ["internal/user/handler.go"],
  "failed_hunks": {},
  "rounds_used": 1
}
On failure: 
{
  "task_id": "verify",
  "status": "FAIL",
  "applied_hunks": [],
  "failed_hunks": {
    "internal/user/handler.go": "syntax: unexpected token '}' at line 42"
  },
  "rounds_used": 1
}
An audit log entry is appended to <log-dir>/kode.log after every run regardless of verdict. Use kode stats to aggregate and trend these entries.

​
Alias

kode verify-hunks is a full alias for kode verify. It accepts the same flags and behaves identically. It exists for compatibility with scripts that used the earlier command name. 
# These are identical
kode verify --input patches.json
kode verify-hunks --input patches.json

​
Example: manual patch-and-verify workflow

# 1. Generate patches without applying them
kode generate "add null check to CreateUser" > hunks.json

# 2. Inspect the JSON before verifying
cat hunks.json | jq '.[] | .file_path'

# 3. Wrap in the verify input envelope (generate outputs a raw array)
echo "{\"hunks\": $(cat hunks.json), \"original_files\": {}}" > patches.json

# 4. Run verification
kode verify --input patches.json

# 5. If it passes, apply with kode generate --apply
kode verify does not run your test suite. It validates structural properties of the code (syntax, imports, call graph, architecture, security). For end-to-end validation including test execution, use kode loop.
For an explanation of each gate and how to interpret failures, see the gate failure troubleshooting guide. For a conceptual overview of the pipeline, see the verification pipeline reference.