grok-cli Test Suite

This directory contains test scripts for validating grok-cli features.

Test Scripts

temperature-feature-test.zsh

Purpose: Automated functional testing of temperature configuration feature Coverage: CLI temperature flags, validation, default behavior, persistence

Usage:

cd /Volumes/DM 2T/Source/Agents/grok-cli
zsh tests/temperature-feature-test.zsh

Test Cases:

• Valid temperature values (0.0, 0.7, 5.0)
• Invalid temperature rejection (-0.1, 5.1, non-numeric)
• Default temperature behavior (0.7)
• State persistence validation

test-venice.js

Purpose: Validate Venice AI backend connection (bug fix verification) Coverage: Tests that Venice AI no longer receives unsupported 'think' parameter

Usage:

cd /Volumes/DM 2T/Source/Agents/grok-cli
bun tests/test-venice.js

Requirements:

• VENICE_API_KEY environment variable must be set
• Project must be built (bun run build)

Test Cases:

• Venice AI connection without HTTP 400 errors
• Basic message exchange with Venice backend

test-duplicate-json.ts

Purpose: Unit test for duplicate JSON handling in tool call arguments (bug fix verification) Coverage: Tests that concatenated/duplicated JSON objects are properly handled

Usage:

cd /Volumes/DM 2T/Source/Agents/grok-cli
bun tests/test-duplicate-json.ts

Requirements:

• Project must be built (bun run build)

Test Cases:

• Exact duplicate objects
• Different duplicate objects
• Triple duplication
• Nested objects duplicated
• Arrays in duplicated objects
• Single valid JSON (regression test)
• Empty object duplicated

test-slash-commands-headless.zsh

Purpose: Automated functional testing of slash commands in headless and no-ink modes (bug fix verification) Coverage: Tests that slash commands work correctly in both -p (headless) and --no-ink (interactive plain) modes

Usage:

cd /Volumes/DM 2T/Source/Agents/grok-cli
zsh tests/test-slash-commands-headless.zsh

Requirements:

• Project must be built (bun run build)

Test Cases:

• /help command in headless mode
• /introspect command in headless mode
• /context command in headless mode
• /persona command in headless mode
• /mood command in headless mode
• /clear command in headless mode
• /models with argument in headless mode
• /models without argument (error expected)
• /context view in headless mode (error expected)
• /context edit in headless mode (error expected)

test-tool-validation.md

Purpose: Documentation of tool argument validation feature (bug fix documentation) Coverage: Explains how tool arguments are validated against schemas to prevent Ollama 400 errors

Context:

• When LLMs return duplicate JSON tool arguments, the duplicate-JSON fix extracts the first object
• If the tool was called with wrong parameters (e.g., getAvailablePersonas with a persona parameter), Ollama rejects with 400 invalid tool call arguments
• Tool validation catches these errors before sending to Ollama, providing clear feedback to the LLM

Implementation:

• validateToolArguments() method validates against tool schemas
• Checks for: no-parameter tools with arguments, unknown parameters, missing required parameters
• Returns helpful error messages instead of Ollama 400 errors

Running Tests

Tests use mock settings to avoid API dependencies and should be executed in the project root directory.

Test Results

Test output includes pass/fail indicators and a summary of results.

Adding New Tests

Add new test scripts to this directory with descriptive names following the pattern: feature-name-test.zsh