terraform-ingest-configure

name	terraform-ingest-configure
description	Builds and validates terraform-ingest config.yaml: repositories, output paths, embeddings, and MCP settings. Use when the user wants to configure terraform-ingest, add repositories, edit config.yaml, or enable vector search.
disable-model-invocation	true

Terraform-Ingest Configure

Guide the user through config.yaml using CLI commands. Ask one question at a time.

Progress checklist

Task Progress:
- [ ] Step 1: Confirm config file path
- [ ] Step 2: Set output and clone directories
- [ ] Step 3: Add repositories
- [ ] Step 4: Configure embeddings (optional)
- [ ] Step 5: Configure MCP behavior (optional)
- [ ] Step 6: Validate configuration
- [ ] Step 7: Hand off to terraform-ingest-ingest

Step 1: Config file path

Default: config.yaml in the project root. Override with --config / -c on all commands.

terraform-ingest config get --target output_dir

Step 2: Directories

terraform-ingest config set --target output_dir --value ./output
terraform-ingest config set --target clone_dir --value ./repos

Use absolute paths in production or container deployments.

Step 3: Add repositories

Ask the user for each repo:

1. Git URL (HTTPS or SSH)
2. Branches to track (e.g. main) — empty list ingests default branch
3. Include tags? How many? (max_tags)
4. Module path (path, default .)
5. Recursive scan? (--recursive for monorepos)

Single repo:

terraform-ingest config add-repo \
  --url https://github.com/terraform-aws-modules/terraform-aws-vpc \
  --name aws-vpc \
  --branches main \
  --max-tags 3

Monorepo with submodules:

terraform-ingest config add-repo \
  --url [e-mail]/myorg/terraform-modules.git \
  --name my-modules \
  --branches main \
  --recursive \
  --path modules

Bulk import from GitHub org (many repos):

export GITHUB_TOKEN=ghp_xxx   # if needed
terraform-ingest import github --org myorg --terraform-only

Verify:

terraform-ingest config get --target repositories

Step 4: Embeddings (optional)

Enable for search_modules_vector in MCP:

terraform-ingest config set --target embedding.enabled --value true
terraform-ingest config set --target embedding.strategy --value chromadb-default
terraform-ingest install-deps --strategy chromadb-default

Strategies: chromadb-default (local, no API key), sentence-transformers, openai, claude.

Avoid chromadb_path: ./chromadb at the project root — it shadows the chromadb Python package. Prefer ./data/chromadb or an absolute path outside the repo.

For OpenAI:

terraform-ingest config set --target embedding.strategy --value openai
# Set OPENAI_API_KEY in environment

Step 5: MCP behavior (optional)

Tune for agents that will connect via MCP later:

terraform-ingest config set --target mcp.ingest_on_startup --value true
terraform-ingest config set --target mcp.auto_ingest --value true
terraform-ingest config set --target mcp.refresh_interval_hours --value 24

Customize agent behavior via mcp.instructions in config — see MCP docs. Do not paste full instructions into chat; edit the YAML field or use config set for nested paths.

Step 6: Validate

Run before ingestion:

terraform-ingest skills validate-config config.yaml

Fix any reported errors. An empty repositories list warns but does not fail — confirm with the user before ingesting.

Step 7: Hand off

Invoke terraform-ingest-ingest to run the first ingestion.

Wizard questions (ask one at a time)

1. Which cloud provider or module sources? (URLs or GitHub org)
2. Public or private repositories?
3. Track branches, tags, or both?
4. Need semantic/vector search for module discovery?
5. Will an AI agent connect via MCP after ingest?

See also

• reference.md — full config schema
• Config CLI docs
• Import command