Configuration

wdpkr resolves every setting through four layers, each overriding the one before it:

defaults  →  config file  →  environment variables  →  CLI flags

Run config list at any time to see the effective value of every setting and exactly where it came from:

wdpkr config list

Interactive setup

The fastest way to get configured is the guided walkthrough — it picks providers and stores your API keys:

wdpkr config init

To set or read individual values:

wdpkr config get embedder.model
wdpkr config set indexer.concurrency 16

The config file

Local settings live in ~/.config/wdpkr/config.yaml. Every field is optional, and environment variables override anything set here.

store:
  provider: turbopuffer
  # Provider-specific settings live in nested blocks:
  # turbopuffer:
  #   api_key: ...                # prefer the TURBOPUFFER_API_KEY env var
  # nidus:                         # local, file-backed, pure-Rust store (provider: nidus)
  #   path: ~/.local/share/wdpkr/nidus          # store directory; or WDPKR_NIDUS_PATH env var

embedder:
  provider: voyage
  model: voyage-code-3
  batch_size: 64
  # ollama_host: http://localhost:11434

summarizer:
  provider: anthropic
  model: claude-haiku-4-5-20251001

indexer:
  # namespace: ""             # auto-derived from git remote if empty
  default_branch: main
  concurrency: 8
  max_cost: 50
  hwm_success_threshold: 0.95

Caution

Keep API keys out of the config file — set them via environment variables (or config set, which stores them per-provider). Credentials in a checked-in or shared file are a liability.

Settings reference

Store

Key	Default	Description
store.provider	turbopuffer	Vector store backend — turbopuffer or nidus
store.turbopuffer.api_key	—	Turbopuffer API key (prefer the env var)
store.nidus.path	~/.local/share/wdpkr/nidus	nidus store directory (local backend)

See Storage for the trade-offs between backends. The flat store.turbopuffer_api_key key is still read for backwards compatibility, but prefer the nested store.turbopuffer.api_key.

Embedder

Key	Default	Description
embedder.provider	voyage	voyage, openai, or ollama
embedder.model	voyage-code-3	Provider-derived if unset
embedder.batch_size	64	Embeddings per request
embedder.ollama_host	http://localhost:11434	Ollama endpoint (local provider)
embedder.voyage_api_key	—	Voyage API key (prefer the env var)
embedder.openai_api_key	—	OpenAI API key (prefer the env var)

Summarizer

Key	Default	Description
summarizer.provider	anthropic	Summarization provider
summarizer.model	claude-haiku-4-5-20251001	Summarization model
summarizer.anthropic_api_key	—	Anthropic API key (prefer the env var)

Indexer

Key	Default	Description
indexer.namespace	(auto)	Index namespace; derived from the git remote when empty
indexer.default_branch	main	Branch the index tracks
indexer.git_remote	(auto)	Remote used to derive the namespace
indexer.concurrency	8	Parallel summarize/embed workers
indexer.max_cost	50	Hard cap (USD) on a single indexing run
indexer.hwm_success_threshold	0.95	Min success rate to advance the indexed high-water mark

Taps

Data sources to index. Omit taps to index repository code only (the default files tap). Each entry has a name and optional per-tap settings. See Taps for the full Linear tap reference.

taps:
  - name: files
  - name: linear
    settings:
      amount: 100
      order_by: updatedAt
      include_comments: true

Key	Default	Description
taps[].name	files	Tap name — files or linear
taps[].settings	—	Per-tap settings (e.g. the Linear tap’s amount, order_by, team, include_comments, api_key_env)

Set the Linear API key via the LINEAR_API_KEY environment variable, not the config file.

Environment variables

Every setting has a matching environment variable, which overrides the config file:

# Credentials
ANTHROPIC_API_KEY              # summarization (required)
TURBOPUFFER_API_KEY            # vector storage (required for default store)
VOYAGE_API_KEY                 # embedding (required for default provider)
OPENAI_API_KEY                 # embedding (OpenAI provider)
OLLAMA_HOST                    # embedding (Ollama provider)
LINEAR_API_KEY                 # Linear tap (when the linear tap is configured)

# Providers & models
WDPKR_STORE_PROVIDER           # turbopuffer | nidus
WDPKR_NIDUS_PATH               # nidus store directory (nidus store)
WDPKR_EMBED_PROVIDER           # voyage | openai | ollama
WDPKR_EMBED_MODEL
WDPKR_EMBED_BATCH_SIZE
WDPKR_SUMMARIZER_PROVIDER
WDPKR_SUMMARIZER_MODEL

# Indexing
WDPKR_NAMESPACE                # override the auto-derived namespace
WDPKR_DEFAULT_BRANCH
WDPKR_GIT_REMOTE
WDPKR_CONCURRENCY
WDPKR_MAX_COST
WDPKR_HWM_SUCCESS_THRESHOLD