skills.rest
Military-Veteran-Team-LPT-Realty
Military-Veteran-Team-LPT-RealtyOFFICIAL
Explore all skills by this author →
GitHub

builtin-llm-models

Documents and executes bulk LLM calls against the Manus built-in OpenAI-compatible proxy.

Updated Sep 16, 2026
Search Tags:#prompt engineering#llm#batch processing#structured output#openai-compatible#manus#model catalog
One-click install
npx skills add https://github.com/Military-Veteran-Team-LPT-Realty/mvt-manus-public-skills --skill builtin-llm-models-military-veteran-team-lpt-realty
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: builtin-llm-models
Source: https://github.com/Military-Veteran-Team-LPT-Realty/mvt-manus-public-skills/tree/main/skills/builtin-llm-models
Command: npx skills add https://github.com/Military-Veteran-Team-LPT-Realty/mvt-manus-public-skills --skill builtin-llm-models-military-veteran-team-lpt-realty

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openai, and includes scripts (resource) components.

What problem does it solve? Calling Manus's built-in LLM proxy incorrectly wastes tokens and causes silent failures: model IDs and prices change, each provider family (GPT, Claude, Gemini) requires different thinking/reasoning parameters, and the wrong max-token parameter returns null content. This Skill provides the verified request shapes and a batch driver so bulk LLM jobs work the first time. ## Core Features & Use Cases - Live Model Catalog Discovery: Fetch current model IDs, pricing, and capabilities from the /models endpoint instead of relying on stale hardcoded lists. - Family-Specific Request Shapes: Correct thinking/reasoning syntax for GPT (reasoning.effort), Claude (thinking budget_tokens), and Gemini (default reasoning), plus the max-token rules that prevent truncation and null-content errors. - Concurrent Batch Driver: The bundled llm_batch.py script runs retrying, parallel bulk jobs (summarization, translation, classification, extraction) with optional strict JSON-schema structured output. - Use Case: You need to summarize 500 support tickets. Fetch the live catalog, pick gpt-5-mini, and run llm_batch.py with a prompt template to produce a JSONL file of summaries with token usage per item. ## Quick Start Ask the AI to fetch the live model catalog and then run the llm_batch.py script to summarize each line of your input file with gpt-5-mini into a JSONL output file.

Frequently Asked Questions about builtin-llm-models

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I call the Manus built-in LLM from a Python script?▼

Use the openai Python SDK with OpenAI(), which auto-reads the preconfigured OPENAI_API_KEY and OPENAI_API_BASE environment variables in the sandbox. Then call client.chat.completions.create with a model ID from the live /models catalog.

Which built-in model should I use for bulk summarization or translation?▼

Use gpt-5-mini as the default cheap, fast workhorse for high-volume tasks, dropping to gpt-5-nano for very large volumes with loose quality tolerance. Skip thinking parameters for these tasks since they add cost and latency with little gain.

Why does Gemini return content null with finish_reason length?▼

This happens when you pass max_completion_tokens to Gemini models on the proxy. Switch to max_tokens (for example 16384) or omit the parameter entirely, and the response content returns normally.

How do I enable extended thinking for Claude models?▼

Pass extra_body with thinking set to type enabled and a budget_tokens value, and set max_tokens strictly greater than budget_tokens or the request fails with a 400 error. Note that claude-opus-4-7 uses adaptive thinking and does not require explicit parameters.

Does the built-in LLM proxy support structured JSON output?▼

Yes, all proxy models support OpenAI's response_format JSON Schema specification. You must set strict to true and additionalProperties to false in the schema, then parse the returned message content as JSON.

Can I stream responses from the built-in LLM proxy?▼

The proxy reports supports_streaming as false and the invokeLLM webdev helper does not expose a stream option. To stream you must modify the helper to send stream true and parse the SSE response manually.

Related Skills

Explore More Software Engineering Skills→
bdambrosiobdambrosio
10

split

Split Notes into smaller Notes using delimiters or JSON structure.

Community
Basic
PatientVibesPatientVibes

human-review-loop

Route low-confidence agent outputs to asynchronous human review queues.

Community
Advanced
davila7davila7
30.5k

ai-product

Guides LLM integration, RAG architecture, and prompt engineering for production AI features.

Community
Intermediate
ebecerra-developerebecerra-developer

chatbot-system

Operates a multi-tenant Groq chatbot SaaS with Sanity config sync and Supabase persistence.

Community
Advanced