anyq

Query any config file in plain English. No jq required.

anyq lets you query JSON, YAML, and TOML files using natural language — powered by your choice of AI provider. Describe what you want to extract; anyq writes the query and runs it. It also works as a conventional jq/yq/tomlq replacement for those who prefer writing expressions directly.

---

AI-Powered Querying

The flagship feature. Instead of learning jq syntax, just describe what you want:

# Set your AI provider once
export ANYQ_AI_PROVIDER=openai   # or: anthropic, gemini, ollama
export OPENAI_API_KEY=sk-...

# Then query in plain English
anyq --ai "names of all admin users" users.json
anyq --ai "which services are disabled" config.yaml
anyq --ai "list all dev dependencies" package.json
anyq --ai "database host and port" app.toml

anyq extracts the schema of your file (key names and types only — no actual values leave your machine), asks the AI to write a jq expression, validates it compiles, then runs it locally against your real data.

your file ──► schema extraction ──► AI (keys + types only) ──► jq expression
                                                                      │
your file ◄────────────────────────────────────────────── run locally ┘

Preview the generated query

Use --show-query to inspect (and optionally cancel) the generated expression before it runs:

anyq --ai "count of pending orders" --show-query orders.json

✦ Generated query: [.orders[] | select(.status == "pending")] | length
Run? [Y/n]:

Supported AI providers

Provider	Env var	Default model
openai (default)	OPENAI_API_KEY	gpt-4o-mini
anthropic	ANTHROPIC_API_KEY	claude-3-5-haiku-20241022
gemini	GEMINI_API_KEY	gemini-2.0-flash
ollama (local, free)	OLLAMA_BASE_URL (default: http://localhost:11434)	qwen2.5-coder

Override the model for any provider with ANYQ_AI_MODEL:

ANYQ_AI_MODEL=gpt-4o anyq --ai "summarize the pipeline stages" .github/workflows/ci.yml

Privacy

anyq never sends your file's values to the AI. It only sends a structural schema like:

{
  "users": [{ "name": "string", "age": "number", "role": "string" }]
}

Sensitive field names (password, token, secret, key, etc.) are automatically replaced with "<redacted>".

---

Installation

Pre-built binary (recommended)

# Linux / macOS
curl -sSfL \
  https://github.com/troll-warlord/anyq/releases/latest/download/anyq_linux_amd64.tar.gz \
  | tar -xz anyq && sudo mv anyq /usr/local/bin/

Homebrew (macOS / Linux)

brew install troll-warlord/tap/anyq

Go install

go install github.com/troll-warlord/anyq@latest

Docker

docker run --rm ghcr.io/troll-warlord/anyq:latest --ai "database host" < config.yaml

---

Usage

anyq [flags] [jq-expression] [file...]
anyq diff [flags] <file1> <file2>
anyq validate [flags] --schema <schema.json> <file>

All flags

Flag	Short	Description
--ai		Natural language query — translated to jq by AI
--show-query		Print the AI-generated query and prompt before running
--input-format	-f	Force input format: json, yaml, toml
--output-format	-o	Output format: json, yaml, toml (default: same as input)
--pretty		Pretty-print output (default: true)
--raw-output	-r	Print strings without surrounding quotes
--compact	-c	Compact JSON output (no whitespace)
--slurp		Read all inputs into an array, then run expression once
--null-input	-n	Use null as input; no file is read
--exit-status	-e	Exit 1 if the last output is false or null
--write-output	-w	Write output to a file instead of stdout
--no-color		Disable syntax highlighting

---

More Examples

Direct jq expressions (no AI needed)

If you already know jq, you can use anyq as a drop-in replacement for jq, yq, and tomlq — just point it at any file format:

# Works on JSON, YAML, and TOML without any flags
anyq '.database.host' config.yaml
anyq '.users[] | select(.age > 30)' users.json
anyq '.dependencies | keys' Cargo.toml
anyq '.tool.poetry.version' pyproject.toml

Format conversion

anyq -o json '.' config.yaml      # YAML → JSON
anyq -o yaml '.' Cargo.toml       # TOML → YAML
anyq -o toml '.' package.json     # JSON → TOML

Piping from stdin

cat config.yaml | anyq --ai "what is the log level"
kubectl get pods -o json | anyq --ai "names of pods not in Running state"

Slurp mode — combine multiple documents

--slurp reads all input documents into a single array before running the expression. Essential for operations that need the full dataset: group_by, unique_by, aggregations across files, or the concatenated JSON that tools like go list -json produce.

# Count packages across the whole codebase
go list -json ./... | anyq --slurp 'length'

# Collect all unique imports across every package
go list -json ./... | anyq --slurp '[.[].Imports // [] | .[]] | unique | sort | .[]' -r

# Combine multiple JSON files and group by a field
anyq --slurp 'group_by(.status)' a.json b.json c.json

# Sum a field across files
anyq --slurp '[.[].score] | add' results-jan.json results-feb.json

# Multi-document YAML (--- separated)
anyq --slurp 'length' deployments.yaml

Semantic diff

Compare two config files regardless of format or key order:

anyq diff old.yaml new.yaml
anyq diff config.yaml config.json          # cross-format
anyq diff --exit-status baseline.yaml current.yaml   # CI gatekeeper

-  .app.debug                                   true
+  .app.replicas                                3
~  .database.host                               "localhost"  →  "production.db.internal"

Schema validation

anyq validate --schema pod.schema.json deployment.yaml
anyq validate -s openapi.json api-config.json
anyq validate --schema schema.json data.toml && echo "safe to deploy"

✗ Validation failed: 2 error(s)

  • .: missing properties: 'version'
  • .port: must be <= 65535 but found 99999

GitHub Actions

- name: Read config value
  uses: troll-warlord/anyq@v0.1.0
  id: read-config
  with:
    expression: '.deploy.region'
    file: config.yaml
    output-format: json

- name: Use the result
  run: echo "Deploying to ${{ steps.read-config.outputs.result }}"

---

How It Works

Natural language request
        │
        ▼
  ┌─────────────────┐
  │  Schema Extract │  ← keys + types only, values never leave your machine
  └────────┬────────┘
           │ schema JSON
           ▼
  ┌─────────────────┐
  │   AI Provider   │  ← OpenAI / Anthropic / Gemini / Ollama
  └────────┬────────┘
           │ jq expression
           ▼
  ┌─────────────────┐
  │  gojq Validate  │  ← compile check; retries with error feedback if invalid
  └────────┬────────┘
           │ validated expression
           ▼
  ┌─────────────────┐        ┌──────────────┐
  │   jq Engine     │ ◄──────│  Real data   │  ← your actual file (local only)
  └────────┬────────┘        └──────────────┘
           │ results
           ▼
  ┌─────────────────┐
  │   Serializer    │  ← JSON / YAML / TOML, syntax-highlighted
  └────────┬────────┘
           │
           ▼
        Output

---

Roadmap

• AI-assisted querying — describe what you want in natural language, get results instantly
• Semantic diff — anyq diff old.yaml new.yaml — format-agnostic, key-order-insensitive
• Schema validation — anyq validate --schema k8s.json deploy.yaml — JSON Schema drafts 4–2020-12
• Syntax highlighting — colorized output in the terminal
• Multiple document streams — --- separated YAML documents and concatenated JSON (e.g. go list -json ./...) via --slurp
• In-place editing (-i flag, like sed -i) — modify files while preserving YAML comments
• Interactive TUI — anyq explore file.yaml — navigable tree view with path extraction
• XML support — extend auto-detection and engine to XML
• Shell completion — anyq completion bash|zsh|fish|powershell

---

Contributing

See CONTRIBUTING.md. Issues and PRs are welcome.

---

License

MIT © 2026 troll-warlord