config set

What It Does

config set updates one key in .ragit/config.toml using the repository-aware config loader and writer.

When To Use / When Not To Use

When to use it

You want to change output, ingest, storage, or embedding settings.
You want a scriptable way to change config without editing TOML by hand.

When not to use it

You only want to inspect current values. Use status.
You want to initialize missing defaults. Use init.

Syntax

pnpm ragit config set <key> <value> [--cwd <path>]

Arguments And Options

<key>: The dotted config key to update.
<value>: The value to store.
--cwd <path>: Run against another repository.

Input And Output Contract

Input is positional only.
Output is a plain stdout confirmation message. This command does not currently expose a JSON envelope.

Examples

Human terminal flow:

pnpm ragit config set output.format markdown

Agent-oriented flow:

pnpm ragit config set ingest.supported_types '["adr","spec"]'

Embedding Provider Setup

RAGit recognizes exactly these provider/model profiles:

Provider	Model	Dimensions
`openai`	`text-embedding-3-small`	1536
`openai`	`text-embedding-3-large`	3072
`ollama`	`nomic-embed-text`	768
`ollama`	`mxbai-embed-large`	1024

Recognition does not make a profile evidence-backed or production-supported. The deterministic local-placeholder/placeholder-v1 profile has 64 dimensions and is for offline development/regression only; its benchmark reports are developmentOnly: true, not production-quality evidence.

For this release, only loopback Ollama nomic-embed-text is production-supported. The OpenAI profiles are recognized and mock-contract-tested, but remain opt-in integration targets without authorized live evidence and are not production-supported by this release.

For OpenAI, make OPENAI_API_KEY available in the command environment before ingest or migration. It is environment-only: RAGit has no API-key config key or CLI argument, and credentials must not be committed or persisted. Configure the initial OpenAI target as follows:

pnpm ragit config set embedding.provider openai
pnpm ragit config set embedding.model text-embedding-3-small

The credential-free default OpenAI provider root is https://api.openai.com; RAGit appends /v1/embeddings. A compatible custom gateway can set embedding.base_url, but this value is a provider root, not the final embeddings endpoint. It must be an absolute http or https URL with no username, password, query, or fragment; trailing slashes are normalized away. Never put a credential in embedding.base_url.

For loopback Ollama, first make the exact model available locally:

ollama pull nomic-embed-text
pnpm ragit config set embedding.provider ollama
pnpm ragit config set embedding.model nomic-embed-text
pnpm ragit config set embedding.base_url http://127.0.0.1:11434

RAGit appends /api/embed for Ollama. Only localhost, 127.0.0.1, and ::1 roots are reported as ollama-local; any non-loopback root is reported as custom.

Failures And Cautions

Invalid keys or invalid values fail before the config file is written.
For real providers, use embedding.provider, embedding.model, embedding.base_url, embedding.timeout_ms, embedding.cache_enabled, and embedding.cache_dir as the supported knobs. embedding.dimensions and embedding.version are legacy-only and only remain meaningful for local-placeholder.
Changing the target embedding provider does not rewrite the persisted store contract by itself. Follow up with migrate embeddings when status reports embedding.needsMigration=true.
A requested provider never silently falls back to local-placeholder. Inspect status, then follow the migration preview/apply flow after changing provider or model.
embedding.cache_dir must stay inside the repository root. If it points outside the repo, cache writes are disabled and doctor reports the config as invalid.
Because output is text-only, agents should validate follow-up state with status.

config set

On this page