Pi - LLM Temperature Override Extension

A pi extension allowing to modify temperature of LLMs via --temperature and /temperature.

It is tested with Gemini models (all I tested it with) and with Claude models.

Limitations

Gemini models work with temperature [0..2] but 0 is not recomended for 3+ family especially for "thinking".

Claude family temperature does NOT work with thinking at all, you have to have --thinking=off with temperature other than 1.

Rest of the README is generated by LLM, take it with a grain of salt

Supports 11 API types, 22 providers, and 343 models out of the box — covering every built-in pi provider.

Features

1. CLI Flag: `--temperature`

Set a global temperature override for the session:

pi -e ./pi-extension-of-temperature --temperature 0.3

Invalid values cause a hard error (stderr + exit 1 in print mode, notification in interactive):

# Exits with error: out of range [0, 2]
pi -e ./pi-extension-of-temperature --temperature 3.0 -p 'hello'

2. Slash Command: `/temperature`

Control temperature mid-session:

/temperature 0.7      # Set temperature
/temperature           # Show current temperature
/temperature off       # Clear override (use provider default)

Validates against the current model's API range immediately.

3. Skill Frontmatter: `model` with inline temperature

Skills can specify a model with temperature in their YAML frontmatter:

---
name: creative-writer
description: Creative writing with high temperature for more varied output
model: gemini-3-pro-preview(temperature=1.5)
---

When you run /skill:creative-writer, the extension:

Parses the model field from frontmatter
Validates the model is in the allowed list (blocks execution if not)
Validates the temperature is within range for that model's API
Switches to the model and sets the temperature
Returns { action: "continue" } so pi still does normal skill expansion

Supported APIs & Temperature Ranges

All 11 API types supported by pi are covered:

API Type	Temperature Range	Providers Using This API
`anthropic-messages`	0 – 1	Anthropic, Kimi Coding, MiniMax, Vercel AI Gateway, OpenCode
`anthropic-vertex`	0 – 1	Anthropic Vertex (external extension)
`openai-responses`	0 – 2	OpenAI
`openai-completions`	0 – 2	xAI (Grok), Groq, Cerebras, HuggingFace, ZAI, OpenRouter, OpenCode
`openai-codex-responses`	0 – 2	OpenAI Codex
`azure-openai-responses`	0 – 2	Azure OpenAI
`google-generative-ai`	0 – 2	Google AI Studio
`google-vertex`	0 – 2	Google Cloud Vertex AI
`google-gemini-cli`	0 – 2	Google Gemini CLI, Google Antigravity
`mistral-conversations`	0 – 1.5	Mistral
`bedrock-converse-stream`	0 – 1	Amazon Bedrock

The --temperature CLI flag and /temperature command work with any model on a supported API. The allowedModels list in models.json only governs which models can be switched to via skill frontmatter.

Allowed Models (343 total)

Models are listed in models.json grouped by API type.

Anthropic — 29 models (anthropic-messages, range: 0–1)

Model ID
`claude-3-5-haiku-20241022`
`claude-3-5-haiku-latest`
`claude-3-5-sonnet-20240620`
`claude-3-5-sonnet-20241022`
`claude-3-7-sonnet-20250219`
`claude-3-7-sonnet-latest`
`claude-3-haiku-20240307`
`claude-3-opus-20240229`
`claude-3-sonnet-20240229`
`claude-haiku-4-5`
`claude-haiku-4-5-20251001`
`claude-opus-4-0`
`claude-opus-4-1`
`claude-opus-4-1-20250805`
`claude-opus-4-20250514`
`claude-opus-4-5`
`claude-opus-4-5-20251101`
`claude-opus-4-6`
`claude-sonnet-4-0`
`claude-sonnet-4-20250514`
`claude-sonnet-4-5`
`claude-sonnet-4-5-20250929`
`claude-sonnet-4-6`
`k2p5`
`kimi-k2-thinking`
`MiniMax-M2`
`MiniMax-M2.1`
`MiniMax-M2.5`
`MiniMax-M2.5-highspeed`

Anthropic Vertex — 3 models (anthropic-vertex, range: 0–1)

Model ID
`claude-sonnet-4-6@default`
`claude-opus-4-6@default`
`claude-haiku-4-5@20251001`

OpenAI — 38 models (openai-responses, range: 0–2)

Model ID
`codex-mini-latest`
`gpt-4`
`gpt-4-turbo`
`gpt-4.1`
`gpt-4.1-mini`
`gpt-4.1-nano`
`gpt-4o`
`gpt-4o-2024-05-13`
`gpt-4o-2024-08-06`
`gpt-4o-2024-11-20`
`gpt-4o-mini`
`gpt-5`
`gpt-5-chat-latest`
`gpt-5-codex`
`gpt-5-mini`
`gpt-5-nano`
`gpt-5-pro`
`gpt-5.1`
`gpt-5.1-chat-latest`
`gpt-5.1-codex`
`gpt-5.1-codex-max`
`gpt-5.1-codex-mini`
`gpt-5.2`
`gpt-5.2-chat-latest`
`gpt-5.2-codex`
`gpt-5.2-pro`
`gpt-5.3-codex`
`gpt-5.3-codex-spark`
`gpt-5.4`
`gpt-5.4-pro`
`o1`
`o1-pro`
`o3`
`o3-deep-research`
`o3-mini`
`o3-pro`
`o4-mini`
`o4-mini-deep-research`

OpenAI-compatible — 68 models (openai-completions, range: 0–2)

Includes models from xAI (Grok), Groq, Cerebras, HuggingFace, ZAI, and OpenRouter:

Model ID	Source Provider
`grok-2`	xAI
`grok-2-1212`	xAI
`grok-2-latest`	xAI
`grok-2-vision`	xAI
`grok-2-vision-1212`	xAI
`grok-2-vision-latest`	xAI
`grok-3`	xAI
`grok-3-fast`	xAI
`grok-3-fast-latest`	xAI
`grok-3-latest`	xAI
`grok-3-mini`	xAI
`grok-3-mini-fast`	xAI
`grok-3-mini-fast-latest`	xAI
`grok-3-mini-latest`	xAI
`grok-4`	xAI
`grok-4-1-fast`	xAI
`grok-4-1-fast-non-reasoning`	xAI
`grok-4-fast`	xAI
`grok-4-fast-non-reasoning`	xAI
`grok-beta`	xAI
`grok-code-fast-1`	xAI
`grok-vision-beta`	xAI
`deepseek-r1-distill-llama-70b`	Groq
`gemma2-9b-it`	Groq
`llama-3.1-8b-instant`	Groq
`llama-3.3-70b-versatile`	Groq
`llama3-70b-8192`	Groq
`llama3-8b-8192`	Groq
`meta-llama/llama-4-maverick-17b-128e-instruct`	Groq
`meta-llama/llama-4-scout-17b-16e-instruct`	Groq
`mistral-saba-24b`	Groq
`moonshotai/kimi-k2-instruct`	Groq
`moonshotai/kimi-k2-instruct-0905`	Groq
`openai/gpt-oss-120b`	Groq
`openai/gpt-oss-20b`	Groq
`qwen-qwq-32b`	Groq
`qwen/qwen3-32b`	Groq
`gpt-oss-120b`	Cerebras
`llama3.1-8b`	Cerebras
`qwen-3-235b-a22b-instruct-2507`	Cerebras
`zai-glm-4.7`	Cerebras
`glm-4.5`	ZAI
`glm-4.5-air`	ZAI
`glm-4.5-flash`	ZAI
`glm-4.5v`	ZAI
`glm-4.6`	ZAI
`glm-4.6v`	ZAI
`glm-4.7`	ZAI
`glm-4.7-flash`	ZAI
`glm-5`	ZAI
`MiniMaxAI/MiniMax-M2.1`	HuggingFace
`MiniMaxAI/MiniMax-M2.5`	HuggingFace
`Qwen/Qwen3-235B-A22B-Thinking-2507`	HuggingFace
`Qwen/Qwen3-Coder-480B-A35B-Instruct`	HuggingFace
`Qwen/Qwen3-Coder-Next`	HuggingFace
`Qwen/Qwen3-Next-80B-A3B-Instruct`	HuggingFace
`Qwen/Qwen3-Next-80B-A3B-Thinking`	HuggingFace
`Qwen/Qwen3.5-397B-A17B`	HuggingFace
`XiaomiMiMo/MiMo-V2-Flash`	HuggingFace
`deepseek-ai/DeepSeek-R1-0528`	HuggingFace
`deepseek-ai/DeepSeek-V3.2`	HuggingFace
`moonshotai/Kimi-K2-Instruct`	HuggingFace
`moonshotai/Kimi-K2-Instruct-0905`	HuggingFace
`moonshotai/Kimi-K2-Thinking`	HuggingFace
`moonshotai/Kimi-K2.5`	HuggingFace
`zai-org/GLM-4.7`	HuggingFace
`zai-org/GLM-4.7-Flash`	HuggingFace
`zai-org/GLM-5`	HuggingFace

OpenAI Codex — 8 models (openai-codex-responses, range: 0–2)

Model ID
`gpt-5.1`
`gpt-5.1-codex-max`
`gpt-5.1-codex-mini`
`gpt-5.2`
`gpt-5.2-codex`
`gpt-5.3-codex`
`gpt-5.3-codex-spark`
`gpt-5.4`

Azure OpenAI — 38 models (azure-openai-responses, range: 0–2)

Model ID
`codex-mini-latest`
`gpt-4`
`gpt-4-turbo`
`gpt-4.1`
`gpt-4.1-mini`
`gpt-4.1-nano`
`gpt-4o`
`gpt-4o-2024-05-13`
`gpt-4o-2024-08-06`
`gpt-4o-2024-11-20`
`gpt-4o-mini`
`gpt-5`
`gpt-5-chat-latest`
`gpt-5-codex`
`gpt-5-mini`
`gpt-5-nano`
`gpt-5-pro`
`gpt-5.1`
`gpt-5.1-chat-latest`
`gpt-5.1-codex`
`gpt-5.1-codex-max`
`gpt-5.1-codex-mini`
`gpt-5.2`
`gpt-5.2-chat-latest`
`gpt-5.2-codex`
`gpt-5.2-pro`
`gpt-5.3-codex`
`gpt-5.3-codex-spark`
`gpt-5.4`
`gpt-5.4-pro`
`o1`
`o1-pro`
`o3`
`o3-deep-research`
`o3-mini`
`o3-pro`
`o4-mini`
`o4-mini-deep-research`

Google Generative AI — 24 models (google-generative-ai, range: 0–2)

Model ID
`gemini-1.5-flash`
`gemini-1.5-flash-8b`
`gemini-1.5-pro`
`gemini-2.0-flash`
`gemini-2.0-flash-lite`
`gemini-2.5-flash`
`gemini-2.5-flash-lite`
`gemini-2.5-flash-lite-preview-06-17`
`gemini-2.5-flash-lite-preview-09-2025`
`gemini-2.5-flash-preview-04-17`
`gemini-2.5-flash-preview-05-20`
`gemini-2.5-flash-preview-09-2025`
`gemini-2.5-pro`
`gemini-2.5-pro-preview-05-06`
`gemini-2.5-pro-preview-06-05`
`gemini-3-flash-preview`
`gemini-3-pro-preview`
`gemini-3.1-flash-lite-preview`
`gemini-3.1-pro-preview`
`gemini-3.1-pro-preview-customtools`
`gemini-flash-latest`
`gemini-flash-lite-latest`
`gemini-live-2.5-flash`
`gemini-live-2.5-flash-preview-native-audio`

Google Vertex AI — 12 models (google-vertex, range: 0–2)

Model ID
`gemini-1.5-flash`
`gemini-1.5-flash-8b`
`gemini-1.5-pro`
`gemini-2.0-flash`
`gemini-2.0-flash-lite`
`gemini-2.5-flash`
`gemini-2.5-flash-lite`
`gemini-2.5-flash-lite-preview-09-2025`
`gemini-2.5-pro`
`gemini-3-flash-preview`
`gemini-3-pro-preview`
`gemini-3.1-pro-preview`

Google Gemini CLI — 15 models (google-gemini-cli, range: 0–2)

Model ID
`gemini-2.0-flash`
`gemini-2.5-flash`
`gemini-2.5-pro`
`gemini-3-flash-preview`
`gemini-3-pro-preview`
`gemini-3.1-pro-preview`
`claude-opus-4-5-thinking`
`claude-opus-4-6-thinking`
`claude-sonnet-4-5`
`claude-sonnet-4-5-thinking`
`claude-sonnet-4-6`
`gemini-3-flash`
`gemini-3.1-pro-high`
`gemini-3.1-pro-low`
`gpt-oss-120b-medium`

Mistral — 25 models (mistral-conversations, range: 0–1.5)

Model ID
`codestral-latest`
`devstral-2512`
`devstral-medium-2507`
`devstral-medium-latest`
`devstral-small-2505`
`devstral-small-2507`
`labs-devstral-small-2512`
`magistral-medium-latest`
`magistral-small`
`ministral-3b-latest`
`ministral-8b-latest`
`mistral-large-2411`
`mistral-large-2512`
`mistral-large-latest`
`mistral-medium-2505`
`mistral-medium-2508`
`mistral-medium-latest`
`mistral-nemo`
`mistral-small-2506`
`mistral-small-latest`
`open-mistral-7b`
`open-mixtral-8x22b`
`open-mixtral-8x7b`
`pixtral-12b`
`pixtral-large-latest`

Amazon Bedrock — 83 models (bedrock-converse-stream, range: 0–1)

Includes Claude, Llama, Mistral, Nova, Cohere, DeepSeek, Qwen, and more on AWS Bedrock:

Model ID
`amazon.nova-2-lite-v1:0`
`amazon.nova-lite-v1:0`
`amazon.nova-micro-v1:0`
`amazon.nova-premier-v1:0`
`amazon.nova-pro-v1:0`
`amazon.titan-text-express-v1`
`amazon.titan-text-express-v1:0:8k`
`anthropic.claude-3-5-haiku-20241022-v1:0`
`anthropic.claude-3-5-sonnet-20240620-v1:0`
`anthropic.claude-3-5-sonnet-20241022-v2:0`
`anthropic.claude-3-7-sonnet-20250219-v1:0`
`anthropic.claude-3-haiku-20240307-v1:0`
`anthropic.claude-3-opus-20240229-v1:0`
`anthropic.claude-3-sonnet-20240229-v1:0`
`anthropic.claude-haiku-4-5-20251001-v1:0`
`anthropic.claude-opus-4-1-20250805-v1:0`
`anthropic.claude-opus-4-20250514-v1:0`
`anthropic.claude-opus-4-5-20251101-v1:0`
`anthropic.claude-opus-4-6-v1`
`anthropic.claude-sonnet-4-20250514-v1:0`
`anthropic.claude-sonnet-4-5-20250929-v1:0`
`anthropic.claude-sonnet-4-6`
`cohere.command-r-plus-v1:0`
`cohere.command-r-v1:0`
`deepseek.r1-v1:0`
`deepseek.v3-v1:0`
`deepseek.v3.2-v1:0`
`eu.anthropic.claude-haiku-4-5-20251001-v1:0`
`eu.anthropic.claude-opus-4-5-20251101-v1:0`
`eu.anthropic.claude-opus-4-6-v1`
`eu.anthropic.claude-sonnet-4-20250514-v1:0`
`eu.anthropic.claude-sonnet-4-5-20250929-v1:0`
`eu.anthropic.claude-sonnet-4-6`
`global.anthropic.claude-haiku-4-5-20251001-v1:0`
`global.anthropic.claude-opus-4-5-20251101-v1:0`
`global.anthropic.claude-opus-4-6-v1`
`global.anthropic.claude-sonnet-4-20250514-v1:0`
`global.anthropic.claude-sonnet-4-5-20250929-v1:0`
`global.anthropic.claude-sonnet-4-6`
`google.gemma-3-27b-it`
`google.gemma-3-4b-it`
`meta.llama3-1-70b-instruct-v1:0`
`meta.llama3-1-8b-instruct-v1:0`
`meta.llama3-2-11b-instruct-v1:0`
`meta.llama3-2-1b-instruct-v1:0`
`meta.llama3-2-3b-instruct-v1:0`
`meta.llama3-2-90b-instruct-v1:0`
`meta.llama3-3-70b-instruct-v1:0`
`meta.llama4-maverick-17b-instruct-v1:0`
`meta.llama4-scout-17b-instruct-v1:0`
`minimax.minimax-m2`
`minimax.minimax-m2.1`
`mistral.ministral-3-14b-instruct`
`mistral.ministral-3-8b-instruct`
`mistral.mistral-large-2402-v1:0`
`mistral.voxtral-mini-3b-2507`
`mistral.voxtral-small-24b-2507`
`moonshot.kimi-k2-thinking`
`moonshotai.kimi-k2.5`
`nvidia.nemotron-nano-12b-v2`
`nvidia.nemotron-nano-9b-v2`
`openai.gpt-oss-120b-1:0`
`openai.gpt-oss-20b-1:0`
`openai.gpt-oss-safeguard-120b`
`openai.gpt-oss-safeguard-20b`
`qwen.qwen3-235b-a22b-2507-v1:0`
`qwen.qwen3-32b-v1:0`
`qwen.qwen3-coder-30b-a3b-v1:0`
`qwen.qwen3-coder-480b-a35b-v1:0`
`qwen.qwen3-next-80b-a3b`
`qwen.qwen3-vl-235b-a22b`
`us.anthropic.claude-haiku-4-5-20251001-v1:0`
`us.anthropic.claude-opus-4-1-20250805-v1:0`
`us.anthropic.claude-opus-4-20250514-v1:0`
`us.anthropic.claude-opus-4-5-20251101-v1:0`
`us.anthropic.claude-opus-4-6-v1`
`us.anthropic.claude-sonnet-4-20250514-v1:0`
`us.anthropic.claude-sonnet-4-5-20250929-v1:0`
`us.anthropic.claude-sonnet-4-6`
`writer.palmyra-x4-v1:0`
`writer.palmyra-x5-v1:0`
`zai.glm-4.7`
`zai.glm-4.7-flash`

How It Works

Temperature Injection Architecture

The extension uses two mechanisms to inject temperature into all provider API calls:

Load-time wrapping (8 APIs): At extension load, pi.registerProvider() replaces the streamSimple function for each API type with a wrapper that checks activeTemperature and spreads it into the options before delegating to the original pure stream function from @mariozechner/pi-ai:

anthropic-messages      → streamSimpleAnthropic
google-generative-ai    → streamSimpleGoogle
google-vertex           → streamSimpleGoogleVertex
google-gemini-cli       → streamSimpleGoogleGeminiCli
openai-responses        → streamSimpleOpenAIResponses
openai-completions      → streamSimpleOpenAICompletions
azure-openai-responses  → streamSimpleAzureOpenAIResponses
mistral-conversations   → streamSimpleMistral

Late-binding (3 APIs): For providers whose stream functions are not exported from @mariozechner/pi-ai (they use lazy loading or come from external extensions), the extension captures the already-registered stream function from ctx.modelRegistry.registeredProviders at session_start and re-registers a temperature-injecting wrapper:

anthropic-vertex        → captured from pi-anthropic-vertex extension
bedrock-converse-stream → captured from lazy-loaded amazon-bedrock module
openai-codex-responses  → captured from lazy-loaded openai-codex module

Configuration: `models.json`

All model and temperature range data is stored in models.json:

{
  "apiTemperatureRanges": {
    "anthropic-messages": { "min": 0, "max": 1 },
    "openai-responses": { "min": 0, "max": 2 },
    "mistral-conversations": { "min": 0, "max": 1.5 },
    ...
  },
  "allowedModels": [
    { "id": "claude-sonnet-4-6", "api": "anthropic-messages" },
    { "id": "gpt-5", "api": "openai-responses" },
    { "id": "grok-4", "api": "openai-completions" },
    ...
  ]
}

apiTemperatureRanges: Maps API type to valid temperature range. Used for validation.
allowedModels: Models that can be switched to via skill frontmatter. Each entry maps a model ID to its API type.

Validation

CLI flag: Validated on session_start. Out of [0, 2] or non-numeric causes a hard error (stderr + exit 1 in print mode)
/temperature command: Validated against the current model's API range
Skill frontmatter: Validated against the requested model's API range
Model switch (model_select event): Warns if current temperature is out of range for the new model

Testing with `test-interceptor.js`

The test-interceptor.js file is a fetch interceptor that wraps globalThis.fetch to log the full HTTP request payload for all outgoing LLM API calls. This lets you verify that temperature is actually being injected into the wire-level HTTP request, not just set in extension state.

What it does

Saves the original globalThis.fetch
Replaces it with a wrapper that pattern-matches request URLs against known provider endpoints
Logs the full request URL, parsed JSON body, and highlights the temperature value if found
Passes the request through to the original fetch (non-destructive — requests still work normally)

Supported endpoint patterns

The interceptor recognizes 13 provider URL patterns:

Pattern	Provider
`googleapis.com`	Google (Generative AI, Vertex, Gemini CLI)
`api.anthropic.com`	Anthropic
`api.openai.com`	OpenAI
`api.mistral.ai`	Mistral
`bedrock-runtime`	AWS Bedrock
`openai.azure.com`	Azure OpenAI
`api.x.ai`	xAI (Grok)
`api.groq.com`	Groq
`api.cerebras.ai`	Cerebras
`api.together.xyz`	Together AI
`openrouter.ai`	OpenRouter
`huggingface.co`	HuggingFace
`api.minimax.chat`	MiniMax

How to invoke

Method 1: NODE_OPTIONS with --import (recommended for ESM)

NODE_OPTIONS="--import ./test-interceptor.js" pi -e . --temperature 0.5 -p 'hello world'

Method 2: NODE_OPTIONS with --require (CJS fallback)

NODE_OPTIONS="--require ./test-interceptor.js" pi -e . --temperature 0.5 -p 'hello world'

Method 3: Temporary import in index.ts (development only)

Add at the top of index.ts:

import './test-interceptor.js';

Then run normally:

pi -e . --temperature 0.5 -p 'hello world'

⚠️ Remove the import before committing. This method is only for local debugging.

Reading the output

When a request is intercepted, you'll see output like:

=== OUTGOING LLM REQUEST ===
Provider: google
URL: https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-preview:streamGenerateContent
PAYLOAD: {
  "contents": [...],
  "generationConfig": {
    "temperature": 0.5,
    "maxOutputTokens": 16384
  }
}

>>> TEMPERATURE FOUND: 0.5 <<<

The >>> TEMPERATURE FOUND: 0.5 <<< line confirms the temperature was successfully injected. If you see >>> NO TEMPERATURE IN PAYLOAD <<<, the override is not reaching the HTTP request.

Testing scenarios

# Test Google Gemini with temperature 1.5
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider google --model gemini-3-pro-preview --temperature 1.5 -p 'write a haiku'

# Test Anthropic Claude with temperature 0.1
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider anthropic --model claude-sonnet-4-6 --temperature 0.1 -p 'review this code'

# Test OpenAI GPT with temperature 1.8
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider openai --model gpt-5 --temperature 1.8 -p 'brainstorm ideas'

# Test Mistral with temperature 0.7
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider mistral --model mistral-large-latest --temperature 0.7 -p 'hello'

# Test xAI Grok with temperature 0.3
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider xai --model grok-4 --temperature 0.3 -p 'explain quantum computing'

# Test that invalid temperature is rejected (should error, no network request)
NODE_OPTIONS="--import ./test-interceptor.js" \
  pi -e . --provider anthropic --model claude-sonnet-4-6 --temperature 1.5 -p 'hello'

# Test /temperature command interactively
NODE_OPTIONS="--import ./test-interceptor.js" pi -e .
# Then in the session: /temperature 0.8
# Then send a message and check the logged payload

Verifying temperature absence (control test)

Run without --temperature to confirm the default behavior sends no temperature:

NODE_OPTIONS="--import ./test-interceptor.js" pi -e . -p 'hello'
# Should show: >>> NO TEMPERATURE IN PAYLOAD <<<

Adding New Models

To add models not yet in the allowed list, edit models.json:

{
  "allowedModels": [
    ...existing entries...,
    { "id": "my-new-model", "api": "openai-responses" }
  ]
}

If the model uses a new API type not yet listed in apiTemperatureRanges, add the range too:

{
  "apiTemperatureRanges": {
    ...existing ranges...,
    "my-custom-api": { "min": 0, "max": 2 }
  }
}

For a completely new API type, you'll also need to add a provider wrapper in index.ts — either as a load-time pi.registerProvider() call (if the streamSimple function is exported from @mariozechner/pi-ai) or as a late-binding capture in the session_start handler.

Anthropic Vertex Support

To use temperature with Anthropic Vertex models, install pi-anthropic-vertex alongside this extension:

pi install git:github.com/basnijholt/pi-anthropic-vertex

Then use Vertex model IDs in skill frontmatter:

---
name: vertex-writer
description: Creative writing via Anthropic Vertex
model: claude-sonnet-4-6@default(temperature=0.8)
---

Or set temperature globally when using a Vertex model:

pi -e ./pi-extension-of-temperature --provider anthropic-vertex --model claude-sonnet-4-6@default --temperature 0.5

Installation

Quick test

pi -e ./pi-extension-of-temperature

Global installation

cp -r pi-extension-of-temperature ~/.pi/agent/extensions/temperature

Project-local

cp -r pi-extension-of-temperature .pi/extensions/temperature

Status Indicator

When a temperature override is active, the footer shows temp:0.5 (or whatever value). Clears on /temperature off.

State Persistence

Temperature state is persisted via pi.appendEntry("temperature-state", ...) on each turn start, and restored from session entries on session_start. The CLI flag always takes precedence over persisted state.

Example Skills

See examples/creative-writer/SKILL.md (Gemini, temp=1.5) and examples/precise-reviewer/SKILL.md (Claude, temp=0.1).

Skill examples for new providers

OpenAI GPT-5 — brainstorming with high temperature:

---
name: brainstormer
description: Wild idea generator using GPT-5 at near-max temperature
model: gpt-5(temperature=1.8)
---

Mistral — balanced creative writing:

---
name: mistral-poet
description: Poetry with Mistral at moderate temperature
model: mistral-large-latest(temperature=1.0)
---

xAI Grok — deterministic analysis:

---
name: grok-analyst
description: Precise data analysis with Grok at low temperature
model: grok-4(temperature=0.1)
---

Bedrock Claude — enterprise-grade review:

---
name: bedrock-reviewer
description: Code review via Bedrock Claude at low temperature
model: anthropic.claude-sonnet-4-6(temperature=0.2)
---

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
README.md		README.md
index.ts		index.ts
models.json		models.json
test-interceptor.js		test-interceptor.js

Folders and files

Latest commit

History

Repository files navigation

Pi - LLM Temperature Override Extension

Limitations

Features

1. CLI Flag: --temperature

2. Slash Command: /temperature

3. Skill Frontmatter: model with inline temperature

Supported APIs & Temperature Ranges

Allowed Models (343 total)

How It Works

Temperature Injection Architecture

Configuration: models.json

Validation

Testing with test-interceptor.js

What it does

Supported endpoint patterns

How to invoke

Reading the output

Testing scenarios

Verifying temperature absence (control test)

Adding New Models

Anthropic Vertex Support

Installation

Quick test

Global installation

Project-local

Status Indicator

State Persistence

Example Skills

Skill examples for new providers

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

1. CLI Flag: `--temperature`

2. Slash Command: `/temperature`

3. Skill Frontmatter: `model` with inline temperature

Configuration: `models.json`

Testing with `test-interceptor.js`

Packages