Rob

Speculative execution wrapper for Personal Assistant with predictive caching and model routing

Research a topic using reasoner+critic agents for thorough analysis

Create a plan using predictor+reasoner agents for goal-oriented planning

Set a flag value in JSON output, parsing input as JSON types (boolean, number, object) or string

Extract a field from JSON input using dot notation. Navigates JSON structure to locate and output the specified field path.

Pass input through unchanged for debugging and testing pipelines

Format retrieved memories into a text block suitable for LLM prompt injection. Parses JSON memory array and sorts by similarity distance.

Analyze and score the importance of a message for memory prioritization. Scores text from 0.0 (trivial) to 1.0 (critical) based on factual content, preferences, and explicit memory requests, providing reasoning for the score.

Remove old or low-importance messages from SQLite database. Deletes messages based on age, importance score, or total count criteria. Returns deleted message IDs and count.

Update last_accessed timestamp and increment access_count for message records in SQLite database

Retrieve messages from the PA2 SQLite database. Queries the messages table and returns conversation history filtered by conversation_id, role, or limit.

Save a message to the PA2 SQLite database with conversation tracking. Creates database and schema if needed, assigns unique message ID, and stores message with metadata.

Delete documents from a ChromaDB vector collection by their IDs

Add documents to a ChromaDB vector collection with embeddings. Reads JSON input with id, content, and metadata fields, uses pa-embed tool to obtain embeddings, and stores the document in the specified ChromaDB collection.

Performs semantic search against a ChromaDB vector collection. Parses query from JSON input, queries the collection using embeddings, and returns matching results with similarity distances.

Embed text into vector representations for semantic search using Ollama's embedding models. Outputs embedding vector with metadata.

bug fix (missing default for --notes argument

Fix scrutiny honesty warnings by rewriting a tool's description to align with its prompt keywords, then verify the fix

Validate a CmdForge tool config.yaml against scrutiny rules before publishing

Build single or multi-tool CmdForge pipelines from natural language descriptions, writing config.yaml and README to ~/.cmdforge/ with proper step composition and ToolStep wiring. Runs system commands to generate READMEs, check scrutiny, and auto-fix honesty warnings

Generate or modify artifacts (diagrams, 3D models, code) using AI

Quick code review with actionable suggestions

Extract and explain errors from large log files

Generate unit tests for code with edge cases

Explain error messages and stack traces in plain language

Explain what code does in plain language

Add docstrings to functions and classes

Review only the added/changed code in a diff

Generate a commit message from a git diff

Add inline comments to code explaining what each section does

Render artifact source code to binary formats (PNG, SVG, PDF)

Rewrite text in a different tone while preserving the message. Use --tone to specify style (e.g., professional, casual, formal, friendly, assertive).

Summarize any text into a concise one-line TL;DR. Perfect for quickly understanding long documents or articles.

Rewrite complex text using simpler words and shorter sentences. Use --level to target a specific reading level (default: 5th grade).

Draft email replies with customizable tone. Pipe an email and get a draft response. Use --tone to set the style (default: professional and friendly).

Expand bullet points and notes into flowing, well-written paragraphs. Maintains the original meaning while adding context and transitions.

Correct grammar, spelling, and punctuation errors in text. Returns only the corrected text without explanations.

Explain complex topics in simple terms a 5-year-old could understand. Uses everyday analogies and avoids jargon.

Analyze implementation plans and extract actionable tasks into structured JSON with priorities

Added README with usage examples and format details

Audit project goals by examining milestones and codebase evidence.

Record from microphone and transcribe to text (convenience wrapper)

Transcribe audio file to text (supports Google Cloud, OpenAI-compatible Whisper, remote-faster-whisper)

Record audio from microphone for a fixed duration

Convert text to speech and play it immediately (auto-chunks long text)

Convert text to speech (supports Google Cloud, Piper, OpenAI-compatible backends)

Generate and validate Python code

Generate changelog from git commits

AI-powered secondary review of scrutiny warnings to identify false positives and provide deeper analysis. Evaluates keyword-triggered warnings in context to determine if they represent actual issues or incidental word usage.