Configuration

config.toml reference covering clipboard, autocomplete, and AI.

Location

Path: ~/.config/jiq/config.toml (all platforms). Optional — defaults are reasonable; only AI requires config.

Clipboard

[clipboard]
backend = "auto"

Value	Use when
auto (default)	OS clipboard first, OSC 52 fallback.
system	Local desktop only.
osc52	SSH / tmux / mosh on a terminal that forwards OSC 52 (Ghostty, kitty, WezTerm, foot).

Theme

[theme]
mode = "auto"

Value	Effect
auto (default)	Detect the terminal background at startup and pick the light or dark palette. Falls back to dark if detection is unavailable (e.g. inside tmux/screen without passthrough, or a non-responding terminal).
light	Force the light palette.
dark	Force the dark palette (the classic Galaxy theme).

Autocomplete

[autocomplete]
array_sample_size = 10

For heterogeneous arrays, increase to union more keys (range: 1–1000).

Tooltip

[tooltip]
auto_show = true

When true (default), the function tooltip appears automatically as the cursor lands on a known jq function. Set to false to require Ctrl+I to open it.

AI

[ai]
enabled              = true
provider             = "anthropic"   # "openai" | "gemini" | "bedrock"
max_context_length   = 100000        # characters of schema/sample context
extra_instructions   = ""            # optional guidance appended to every AI prompt
request_timeout_secs = 120           # whole-request timeout; 0 disables it

Provider	Recommended model
Anthropic	claude-haiku-4-5-20251001
OpenAI	gpt-4o-mini
Gemini	gemini-3-flash
Bedrock	global.anthropic.claude-haiku-4-5-20251001-v1:0

Every provider takes an optional effort level for reasoning-capable models; Anthropic and Bedrock Claude also take a context_1m toggle:

[ai.anthropic]
api_key    = "sk-ant-..."
model      = "claude-sonnet-4-6"
effort     = "high"    # low | medium | high | xhigh | max (Claude 4.6+)
context_1m = false     # 1M-token context window beta (Claude Sonnet 4/4.5)

[ai.bedrock]
region     = "us-east-1"
model      = "global.anthropic.claude-sonnet-4-6-v1:0"   # or "us.openai.gpt-5.6-sol"
effort     = "high"    # shaped per model family (Claude vs OpenAI)
context_1m = false     # Claude Sonnet 4/4.5 only

[ai.openai]
api_key = "sk-proj-..."
model   = "gpt-5.1-mini"
effort  = "medium"     # minimal | low | medium | high | xhigh | max

[ai.gemini]
api_key = "AIza..."
model   = "gemini-3-flash"
effort  = "medium"     # minimal | low | medium | high (Gemini 3+)

OpenAI models on Bedrock work through provider = "bedrock" (Converse + AWS profile) or the OpenAI-compatible endpoint with a Bedrock API key. See AI assistant for details per provider.

See AI assistant for per-provider config.

Full example

[clipboard]
backend = "auto"

[theme]
mode = "auto"

[autocomplete]
array_sample_size = 25

[ai]
enabled            = true
provider           = "anthropic"
max_context_length = 80000

[ai.anthropic]
api_key = "sk-ant-..."
model   = "claude-haiku-4-5-20251001"

Query

[query]
debounce_ms = 150   # delay between the last keystroke and the jq re-run

Lower values feel snappier on small files; raise it if large inputs make retyping laggy.

History

[history]
max_entries = 1000   # cap on persisted query-history entries

Save

[save]
default_pattern = "jiq-{timestamp}.json"   # initial filename in the save dialog

Supports {timestamp}, {cwd}, and ~ expansion, e.g. ~/exports/jiq-{timestamp}.json.

Environment overrides

Variable	Effect
JIQ_DEBUG=1	Same as --debug: write debug logs to /tmp/jiq-debug.log.

See Troubleshooting.