bearing-genai

Plugin bearing-genai from marketplace bearing in Deepta-AI/bearing. Commands read from the repo's manifest — 9 days ago.

Skills (14)

• computer-vision

  Image and scan features: check or classify photos, read fields from scans, find defects. Use when asked to "classify images" or "extract from scans".

• dataset-generation

  Generates eval, training or preference data from a real seed, deduped and audited. Use when asked to "make synthetic data" or "we need more examples".

• genai-design

  Designs a GenAI solution before code: metric, cheaper option, model, cost, risks, evals. Use when asked "RAG or agent" or "design the LLM feature".

• llm-agent

  Builds an LLM agent on the Anthropic SDK: tools, bounded loop, memory, guardrails, kill switch. Use when asked to "build an agent" or "multi-agent".

• llm-eval

  Evaluates an LLM feature: golden set, code graders, LLM judge, CI regression gate. Use when asked to "build an eval set" or "evaluate the model".

• llm-fine-tuning

  Fine-tunes an open model when prompting and RAG fall short: LoRA, DPO, distillation. Use when asked to "fine-tune a model" or "train on our data".

• llm-gateway

  One gateway for every LLM call: find each call, model tiers, retries, caching, cost per feature. Use when asked about "AI spend" or "LLM costs".

• llm-guardrails

  Secures LLM chatbots and agents: prompt injection, jailbreaks, PII in prompts, tool permissions. Use when asked for "guardrails" or "LLM safety".

• mcp-server

  Builds an MCP server in Python or TypeScript with tested tools and auth. Use when asked to "build an MCP server" or "expose this as MCP tools".

• model-routing

  Picks the cheapest LLM model or cascade per feature that still passes its eval. Use when asked "which model should we use" or "cut our LLM bill".

• prompt-registry

  Keeps prompts as versioned files with typed variables, fixture tests and scores. Use when asked to "write the system prompt" or "version prompts".

• rag

  Builds RAG on pgvector: ingestion, chunking, a retriever that cites sources, metrics. Use when asked to "add RAG" or "answer from our documents".

• speech

  Builds transcription, text-to-speech or voice agents, measured by WER and latency. Use when asked to "build a voice agent" or "speech to text".

• tabular-ml

  Predictive models on tables or event history: churn and risk scores, demand forecasts, anomalies. Use when asked to "predict" or "forecast" from data.

Similar plugins