### Run ExLLM Example App Interactively

Source: https://github.com/azmaveth/ex_llm/blob/master/examples/README.md

Start the example application in interactive mode. Ensure Ollama is installed and running.

```bash
# Ensure Ollama is running:
ollama serve  # if not already running

# Run the example app interactively:
./example_app.exs
```

--------------------------------

### Install Go Gemini Library

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/GEMINI-DOCS-03-LIBRARIES.md

Install the recommended Go library for the Gemini API using go get.

```bash
go get google.golang.org/genai
```

--------------------------------

### Install and Serve Ollama

Source: https://github.com/azmaveth/ex_llm/blob/master/TEST_SETUP.md

Installs Ollama using a script and starts the Ollama service. A model is then pulled for testing.

```bash
# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Start Ollama service
ollama serve

# Pull a model for testing
ollama pull llama3.2

# Set environment variable (if not using default port)
export OLLAMA_HOST=http://localhost:11434
```

--------------------------------

### Create SFT Job with All Settings

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Example of starting a supervised fine-tuning job with various optional settings.

```shell
firectl create sftj \
--base-model llama-v3p1-8b-instruct \
--dataset cancerset \
--output-model my-tuned-model \
--job-id my-fine-tuning-job \
--learning-rate 0.0001 \
--epochs 2 \
--early-stop \
--evaluation-dataset my-eval-set
```

--------------------------------

### Advanced Agent Example Setup

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

Provides the initial Python imports for an advanced agent example that combines LM Studio models with custom functions. This setup allows the agent to perform actions like opening URLs, checking time, and analyzing file systems.

```python
import json
from urllib.parse import urlparse
import webbrowser
from datetime import datetime
import os
from openai import OpenAI
```

--------------------------------

### Install and Serve Ollama

Source: https://github.com/azmaveth/ex_llm/blob/master/test/README.md

Installs Ollama locally and starts the Ollama service. This is required for running tests that depend on local LLM services.

```bash
# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Start Ollama service
ollama serve
```

--------------------------------

### Clone and Install ML-Agents Library

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/huggingface-hub/llms.txt

Clone the ML-Agents repository and install the necessary packages. This is the initial setup required to work with ML-Agents.

```bash
# Clone the repository
git clone https://github.com/Unity-Technologies/ml-agents

# Go inside the repository and install the package
cd ml-agents
pip3 install -e ./ml-agents-envs
pip3 install -e ./ml-agents
```

--------------------------------

### GitHub Actions CI/CD Configuration Example

Source: https://github.com/azmaveth/ex_llm/blob/master/test/ENVIRONMENT_SETUP.md

Example environment variable setup for a GitHub Actions workflow, enabling test caching and skipping Bumblebee tests.

```yaml
# Example GitHub Actions setup
env:
  EX_LLM_TEST_CACHE_ENABLED: true
  BUMBLEBEE_SKIP_TESTS: true
  # Add API keys as secrets
```

--------------------------------

### Fal Provider Setup

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the Fal provider using npm, pnpm, or yarn.

```APIDOC
## Installation

### npm
```bash
npm install @ai-sdk/fal
```

### pnpm
```bash
pnpm add @ai-sdk/fal
```

### yarn
```bash
yarn add @ai-sdk/fal
```
```

--------------------------------

### Quick Example: Chat with Llama Model (TypeScript)

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

Demonstrates how to initialize the LMStudioClient, load a Llama model, and get a response to a prompt. Requires the @lmstudio/sdk package.

```typescript
import { LMStudioClient } from "@lmstudio/sdk";
const client = new LMStudioClient();

const model = await client.llm.model("llama-3.2-1b-instruct");
const result = await model.respond("What is the meaning of life?");

console.info(result.content);
```

--------------------------------

### Run OAuth2 Setup Script

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/OAUTH2_SETUP_GUIDE.md

Execute the Elixir script to start a local web server, obtain authorization, and save access/refresh tokens to a file.

```bash
# From the ex_llm directory
elixir scripts/setup_oauth2.exs
```

--------------------------------

### Generate Content with PHP (Guzzle)

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

This PHP example uses the Guzzle HTTP client to send a GET request. Make sure Guzzle is installed via Composer.

```php
<?php

$client = new \GuzzleHttp\Client();

$response = $client->request('GET', 'https://openrouter.ai/api/v1/generation?id=id', [
  'headers' => [
    'Authorization' => 'Bearer <token>',
  ],
]);

echo $response->getBody();
?>
```

--------------------------------

### Example: Create a Deployment

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

An example of creating a deployment for the 'falcon-7b' model.

```bash
firectl create deployment falcon-7b
```

--------------------------------

### Python: Get Text Embeddings with OpenAI Client

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

Example of using the OpenAI Python client to generate text embeddings from LM Studio. Requires the 'openai' library to be installed.

```python
from openai import OpenAI
client = OpenAI(base_url="http://localhost:1234/v1", api_key="lm-studio")

def get_embedding(text, model="model-identifier"):
   text = text.replace("\n", " ")
   return client.embeddings.create(input = [text], model=model).data[0].embedding

print(get_embedding("Once upon a time, there was a cat."))
```

--------------------------------

### Run Reward Kit Examples

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Navigate to the reward-kit directory, activate the virtual environment, and run the cosine_scaled_example.py script to see the reward function in action.

```bash
# Make sure you're in the reward-kit directory
cd /path/to/reward-kit

# Activate the virtual environment
source .venv/bin/activate

# Run the example
python examples/accuracy_length/cosine_scaled_example.py
```

--------------------------------

### Get API Key using PHP (Guzzle)

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

This PHP example uses the Guzzle HTTP client to retrieve your API key. Make sure Guzzle is installed via Composer.

```php
<?php

$client = new \GuzzleHttp\Client();

$response = $client->request('GET', 'https://openrouter.ai/api/v1/key', [
  'headers' => [
    'Authorization' => 'Bearer <token>',
  ],
]);

echo $response->getBody();
?>
```

--------------------------------

### Provide Router Troubleshooting Context

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/GEMINI-DOCS-23-PROMPT-ENGINEERING.md

Include specific troubleshooting guides in the prompt to get customized responses for device issues. This example shows how to provide router LED status information for accurate troubleshooting.

```text
What should I do to fix my disconnected wifi? The light on my Google
Wifi router is yellow and blinking slowly.
```

```text
A yellow blinking light on your Google Wifi router usually indicates a problem
with the internet connection. Here are some steps to troubleshoot and fix the
issue:
1. Restart your router and modem:
- Unplug your modem from the power outlet and wait for 30 seconds.
- Plug your modem back in and wait for it to fully boot up.
- Unplug your Google Wifi router and wait for 30 seconds.
- Plug your Google Wifi router back in and wait for it to fully boot up.
.....
```

```text
Answer the question using the text below. Respond with only the text provided.
Question: What should I do to fix my disconnected wifi? The light on my Google Wifi router is yellow and blinking slowly.

Text:
Color: Slowly pulsing yellow
What it means: There is a network error.
What to do:
Check that the Ethernet cable is connected to both your router and your modem and both devices are turned on. You might need to unplug and plug in each device again.

Color: Fast blinking yellow
What it means: You are holding down the reset button and are factory resetting this device.
What to do:
If you keep holding down the reset button, after about 12 seconds, the light will turn solid yellow. Once it is solid yellow, let go of the factory reset button.

Color: Solid yellow
What it means: Router is factory resetting.
What to do:
This can take up to 10 minutes. When it's done, the device will reset itself and start pulsing white, letting you know it's ready for setup.

Color: Solid red
What it means: Something is wrong.
What to do:
Critical failure. Factory reset the router. If the light stays red, contact Wifi customer support.
```

```text
Check that the Ethernet cable is connected to both your router and your modem and both devices are turned on. You might need to unplug and plug in each device again.
```

--------------------------------

### Example: Create a Dataset

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

An example of creating a dataset named 'my-dataset' from a local JSONL file.

```bash
firectl create dataset my-dataset /path/to/dataset.jsonl
```

--------------------------------

### Install GitHub App via Claude Code CLI

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Use the /install-github-app command in the Claude Code terminal to automate the setup of the GitHub app and required secrets for GitHub Actions integration. This is the recommended quickstart method for direct Anthropic API users.

```bash
/install-github-app
```

--------------------------------

### Get API Key using Python

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

This Python script uses the requests library to get your API key. Make sure to install the library if you haven't already (`pip install requests`).

```python
import requests

url = "https://openrouter.ai/api/v1/key"

headers = {"Authorization": "Bearer <token>"}

response = requests.get(url, headers=headers)

print(response.json())
```

--------------------------------

### example.py - Create File Tool Usage

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

Demonstrates using the create_file tool with the LM Studio convenience API to generate a file based on an LLM's response.

```python
import lmstudio as lms
from create_file_tool import create_file

model = lms.llm("qwen2.5-7b-instruct")
model.act(
  "Please create a file named output.txt with your understanding of the meaning of life.",
  [create_file],
)

```

--------------------------------

### Install Dependencies

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/QUICKSTART.md

After adding ExLLM to your dependencies, run this command to fetch them.

```bash
mix deps.get
```

--------------------------------

### Completions API Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

This example demonstrates how to make a POST request to the /v1/completions endpoint to get text completions from a model.

```APIDOC
## POST /v1/completions

### Description
This endpoint is used to get text completions from a selected model. It requires a model and a prompt.

### Method
POST

### Endpoint
https://openrouter.ai/api/v1/completions

### Parameters
#### Request Body
- **model** (string) - Required - The model to use for completion.
- **prompt** (string) - Required - The prompt to send to the model.

### Request Example
```json
{
  "model": "model",
  "prompt": "prompt"
}
```

### Response
#### Success Response (200)
- The response body will contain the completion details.
```

--------------------------------

### Install Sample-Factory Library

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/huggingface-hub/llms.txt

Install the Sample-Factory library using pip. This library is known to work on Linux and MacOS, with no Windows support currently available.

```bash
pip install sample-factory
```

--------------------------------

### Direct HTTPClient GET Call

Source: https://github.com/azmaveth/ex_llm/blob/master/http_client_references.md

Example of a direct GET request using the legacy HTTPClient. Found in circuit breaker logic.

```Elixir
HTTPClient.get("/api/data")
```

--------------------------------

### Install Replicate Provider with yarn

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the AI SDK and the Replicate provider using yarn.

```bash
yarn add ai @ai-sdk/replicate
```

--------------------------------

### ExLLM Example App CLI Commands

Source: https://github.com/azmaveth/ex_llm/blob/master/examples/README.md

Execute specific demos or list available options using command-line arguments. Useful for automation or testing specific functionalities.

```bash
# Show help and available demos
elixir example_app.exs --help

# List all available demos
elixir example_app.exs --list

# Run a specific demo
elixir example_app.exs basic-chat

# Run a demo with custom arguments
elixir example_app.exs basic-chat "What is the weather like?"
```

--------------------------------

### Install Anthropic TypeScript SDK

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Install the official Anthropic TypeScript SDK using npm. This is required to use the TypeScript code examples.

```bash
npm install @anthropic-ai/sdk
```

--------------------------------

### Install Anthropic Python SDK

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Install the official Anthropic Python SDK using pip. This is required to use the Python code examples.

```bash
pip install anthropic
```

--------------------------------

### Install Replicate Provider with pnpm

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the AI SDK and the Replicate provider using pnpm.

```bash
pnpm add ai @ai-sdk/replicate
```

--------------------------------

### General Tool Use for User Requests

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

This example shows how to use tools to answer user requests. It includes thinking steps for tool selection and parameter inference. The model will ask for missing parameters if required ones are not provided.

```markdown
Answer the user's request using relevant tools (if they are available). Before calling a tool, do some analysis within <thinking></thinking> tags. First, think about which of the provided tools is the relevant tool to answer the user's request. Second, go through each of the required parameters of the relevant tool and determine if the user has directly provided or given enough information to infer a value. When deciding if the parameter can be inferred, carefully consider all the context to see if it supports a specific value. If all of the required parameters are present or can be reasonably inferred, close the thinking tag and proceed with the tool call. BUT, if one of the values for a required parameter is missing, DO NOT invoke the function (not even with fillers for the missing params) and instead, ask the user to provide the missing parameters. DO NOT ask for more information on optional parameters if it is not provided.
```

--------------------------------

### Task Input Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/GEMINI-DOCS-23-PROMPT-ENGINEERING.md

Provide clear, step-by-step instructions for tasks you want the model to perform. Specify constraints like the number of items.

```text
Give me a simple list of just the things that I must bring on a camping trip. The list should have 5 items.
```

--------------------------------

### Install Replicate Provider with npm

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the AI SDK and the Replicate provider using npm.

```bash
npm install ai @ai-sdk/replicate
```

--------------------------------

### Copying .env Example File

Source: https://github.com/azmaveth/ex_llm/blob/master/test/README.md

Create a local .env file by copying the example configuration file. This file is used to store sensitive API keys and environment-specific settings.

```bash
cp .env.example .env
```

--------------------------------

### Few-Shot Learning Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/GEMINI-DOCS-26-FILES-API.md

Use few-shot examples to guide the model's output format and content. This helps the model understand the desired relationship between inputs and outputs.

```text
Determine the city along with the landmark.
city: Rome, landmark: the Colosseum.
city: Beijing, landmark: Forbidden City
city: Rio de Janeiro, landmark: Christ the Redeemer statue
```

--------------------------------

### Few-shot Prompt Example for Concise Responses

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/gemini/GEMINI-DOCS-23-PROMPT-ENGINEERING.md

A few-shot prompt that includes examples to guide the model towards producing concise responses, demonstrating preference for shorter explanations.

```plaintext
Below are some examples showing a question, explanation, and answer format:

Question: Why is the sky blue?
Explanation1: The sky appears blue because of Rayleigh scattering, which causes
shorter blue wavelengths of light to be scattered more easily than longer red
wavelengths, making the sky look blue.
Explanation2: Due to Rayleigh scattering effect.
Answer: Explanation2

Question: What is the cause of earthquakes?
Explanation1: Sudden release of energy in the Earth's crust.
Explanation2: Earthquakes happen when tectonic plates suddenly slip or break
apart, causing a release of energy that creates seismic waves that can shake the
ground and cause damage.
Answer: Explanation1

Now, Answer the following question given the example formats above:

Question: How is snow formed?
Explanation1: Snow is formed when water vapor in the air freezes into ice
crystals in the atmosphere, which can combine and grow into snowflakes as they
fall through the atmosphere and accumulate on the ground.
Explanation2: Water vapor freezes into ice crystals forming snow.
Answer:

```

--------------------------------

### ExLLM Application Configuration Example

Source: https://github.com/azmaveth/ex_llm/blob/master/MIGRATION_GUIDE_V1.md

Shows how ExLLM application configuration in `config/config.exs` remains compatible with v1.0.0, including setting default providers and provider-specific options.

```elixir
config :ex_llm,
  default_provider: :openai,
  openai: [
    api_key: System.get_env("OPENAI_API_KEY"),
    model: "gpt-4"
  ]
```

--------------------------------

### Full HTTP Stream Response Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

An example of a complete HTTP stream response from the Anthropic API, including message start, content blocks, and message stop events.

```JSON
event: message_start
data: {"type": "message_start", "message": {"id": "msg_1nZdL29xx5MUA1yADyHTEsnR8uuvGzszyY", "type": "message", "role": "assistant", "content": [], "model": "claude-opus-4-20250514", "stop_reason": null, "stop_sequence": null, "usage": {"input_tokens": 25, "output_tokens": 1}}}

event: content_block_start
data: {"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}}

event: ping
data: {"type": "ping"}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "Hello"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "!"}}

event: content_block_stop
data: {"type": "content_block_stop", "index": 0}

event: message_delta
data: {"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence":null}, "usage": {"output_tokens": 15}}

event: message_stop
data: {"type": "message_stop"}
```

--------------------------------

### Test Suite Setup: Environment File

Source: https://github.com/azmaveth/ex_llm/blob/master/ENVIRONMENT.md

Copy the test environment template and edit it to add API keys.

```bash
# 1. Copy the test environment template
cp .env.test.example .env.test

# 2. Edit .env.test and add your API keys
# Note: You don't need ALL keys - tests will skip providers without keys
```

--------------------------------

### Example Computer Use Tool Actions (JSON)

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Demonstrates common actions supported by the computer use tool, including taking screenshots, clicking at coordinates, typing text, and scrolling. These are examples of how the tool can be invoked.

```json
// Take a screenshot
{
  "action": "screenshot"
}
```

```json
// Click at position
{
  "action": "left_click",
  "coordinate": [500, 300]
}
```

```json
// Type text
{
  "action": "type",
  "text": "Hello, world!"
}
```

```json
// Scroll down (Claude 4/3.7)
{
  "action": "scroll",
  "coordinate": [500, 400],
  "scroll_direction": "down",
  "scroll_amount": 3
}
```

--------------------------------

### Perplexity Provider Setup

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the Perplexity provider using npm, pnpm, or yarn.

```APIDOC
## Perplexity Provider Setup

Install the Perplexity provider using:

- **pnpm:** `pnpm add @ai-sdk/perplexity`
- **npm:** `npm install @ai-sdk/perplexity`
- **yarn:** `yarn add @ai-sdk/perplexity`
```

--------------------------------

### Configure ExLLM Example App with Different Providers

Source: https://github.com/azmaveth/ex_llm/blob/master/examples/README.md

Set the PROVIDER environment variable to use different LLM services like OpenAI, Anthropic, Groq, X.AI, or a mock provider for testing. Ensure the necessary API keys are set.

```bash
# Interactive mode examples:
# OpenAI
export OPENAI_API_KEY="your-key"
PROVIDER=openai ./example_app.exs

# Anthropic (Claude)
export ANTHROPIC_API_KEY="your-key"
PROVIDER=anthropic ./example_app.exs

# Groq (Fast Cloud)
export GROQ_API_KEY="your-key"
PROVIDER=groq ./example_app.exs

# X.AI (Grok)
export XAI_API_KEY="your-key"
PROVIDER=xai ./example_app.exs

# Mock (Testing - no API key needed)
PROVIDER=mock ./example_app.exs

# Non-interactive mode examples:
PROVIDER=openai elixir example_app.exs basic-chat "Hello from OpenAI!"
PROVIDER=anthropic elixir example_app.exs function-calling
PROVIDER=mock elixir example_app.exs provider-capabilities
```

--------------------------------

### Initial OAuth2 Setup Script

Source: https://github.com/azmaveth/ex_llm/blob/master/test/OAUTH2_TESTING.md

Command to run the one-time setup script for performing the initial OAuth2 flow. This generates a `.gemini_tokens` file.

```bash
# One-time setup to get initial tokens
elixir scripts/setup_oauth2.exs
```

--------------------------------

### Provide Example Customer Interactions

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Includes sample 'good' customer interactions to guide Claude's response style and content. These examples help define guardrails and expected conversational flow.

```python
EXAMPLES="""
Here are a few examples of how you can interact with customers:

<example 1>
H: Hi, do you offer commercial insurance for small businesses?

A: Ah! Unfortunately, we don't offer commercial insurance at this time.

```

--------------------------------

### Install Adapters Library

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/huggingface-hub/llms.txt

Install the Adapters library using pip. This is the first step to using adapters with Hugging Face transformers.

```bash
pip install adapters
```

--------------------------------

### Example YAML Section for Model Card Metadata

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/huggingface-hub/llms.txt

This is an example of how to structure the YAML section in your README.md file to include various metadata fields. Ensure the file starts and ends with '---' to denote the YAML block.

```yaml
---
language: 
  - "List of ISO 639-1 code for your language"
  - lang1
  - lang2
thumbnail: "url to a thumbnail used in social sharing"
tags:
- tag1
- tag2
license: "any valid license identifier"
datasets:
- dataset1
- dataset2
metrics:
- metric1
- metric2
base_model: "base model Hub identifier"
---
```

--------------------------------

### LM Studio Default Tool Use Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

This example demonstrates the default tool use format provided by LM Studio when a model does not have native tool support. It shows the system prompt, available tools, rules, and an example user query.

```bash
-> % lms log stream
Streaming logs from LM Studio

timestamp: 11/13/2024, 9:35:15 AM
type: llm.prediction.input
modelIdentifier: gemma-2-2b-it
modelPath: lmstudio-community/gemma-2-2b-it-GGUF/gemma-2-2b-it-Q4_K_M.gguf
input: "<start_of_turn>system
You are a tool-calling AI. You can request calls to available tools with this EXACT format:
[TOOL_REQUEST]{"name": "tool_name", "arguments": {"param1": "value1"}}[END_TOOL_REQUEST]

AVAILABLE TOOLS:
{
  "type": "toolArray",
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_delivery_date",
        "description": "Get the delivery date for a customer's order",
        "parameters": {
          "type": "object",
          "properties": {
            "order_id": {
              "type": "string"
            }
          },
          "required": [
            "order_id"
          ]
        }
      }
    }
  ]
}

RULES:
- Only use tools from AVAILABLE TOOLS
- Include all required arguments
- Use one [TOOL_REQUEST] block per tool
- Never use [TOOL_RESULT]
- If you decide to call one or more tools, there should be no other text in your message

Examples:
"Check Paris weather"
[TOOL_REQUEST]{"name": "get_weather", "arguments": {"location": "Paris"}}[END_TOOL_REQUEST]

"Send email to John about meeting and open browser"
[TOOL_REQUEST]{"name": "send_email", "arguments": {"to": "John", "subject": "meeting"}}[END_TOOL_REQUEST]
[TOOL_REQUEST]{"name": "open_browser", "arguments": {}}[END_TOOL_REQUEST]

Respond conversationally if no matching tools exist.<end_of_turn>
<start_of_turn>user
Get me delivery date for order 123<end_of_turn>
<start_of_turn>model
"
```

--------------------------------

### Import Default ElevenLabs Provider

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Import the default provider instance for ElevenLabs. This is the simplest way to get started.

```typescript
import { elevenlabs } from '@ai-sdk/elevenlabs';
```

--------------------------------

### Tool Use Evaluation Setup

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Sets up the schema and expected behavior for evaluating a 'search' tool. This is a common use case for assessing model tool utilization.

```python
import json
from reward_kit.rewards.function_calling import composite_function_call_reward

# Define expected schema for search function
search_schema = {
    "name": "search",
    "arguments": {
        "query": {"type": "string"}
    }
}

expected_behavior = """
The search function should be called:
1. When the user is asking for factual information
2. With a clear, specific query that captures what the user is looking for
3. Without including instructions, formatting requests, or explanations in the query
"""
```

--------------------------------

### Mistral AI Provider Setup

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/vercel/llms.txt

Install the Mistral AI provider package using your preferred package manager.

```APIDOC
## Install Mistral AI Provider

### Description
Installs the Mistral AI provider package.

### Package Manager

- **pnpm**
  ```sh
  pnpm add @ai-sdk/mistral
  ```

- **npm**
  ```sh
  npm install @ai-sdk/mistral
  ```

- **yarn**
  ```sh
  yarn add @ai-sdk/mistral
  ```
```

--------------------------------

### Example: Set Theme Globally

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

This example demonstrates setting the color theme globally. Options are 'dark', 'light', 'light-daltonized', or 'dark-daltonized'.

```bash
claude config set -g theme dark-daltonized
```

--------------------------------

### Set Up OAuth2 for Google/Gemini Tests

Source: https://github.com/azmaveth/ex_llm/blob/master/test/ENVIRONMENT_SETUP.md

Configure environment variables for Google OAuth2 client ID and secret, then run a setup script to authenticate and create token files.

```bash
# 1. Set environment variables
export GOOGLE_CLIENT_ID=your_client_id
export GOOGLE_CLIENT_SECRET=your_client_secret

# 2. Run setup script
elixir scripts/setup_oauth2.exs

# 3. Follow the prompts to authenticate
# This creates .gemini_tokens file
```

--------------------------------

### Complete Request Lifecycle Logging Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/LOGGER.md

An end-to-end example demonstrating how to log the entire lifecycle of a request, from context setup to API calls and error handling. It uses Logger.with_context for managing log context across operations.

```elixir
defmodule MyApp.AI do
  alias ExLLM.Logger
  
  def generate_summary(text, user_id) do
    # Set up context
    Logger.with_context(
      operation: :summary,
      user_id: user_id,
      request_id: UUID.uuid4()
    ) do
      Logger.info("Starting summary generation")
      
      messages = [
        %{role: "system", content: "You are a helpful assistant."},
        %{role: "user", content: "Summarize: #{text}"}
      ]
      
      # The actual API call (logging handled internally)
      case ExLLM.chat(:openai, messages) do
        {:ok, response} ->
          Logger.info("Summary generated",
            tokens: response.usage.total_tokens,
            cost_usd: response.cost.total_cost
          )
          {:ok, response.content}
          
        {:error, reason} ->
          Logger.error("Summary generation failed", error: reason)
          {:error, reason}
      end
    end
  end
end
```

--------------------------------

### Install Claude Code with User-Writable npm Prefix

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

This script guides you through setting up a dedicated directory for global npm packages, preventing permission errors and ensuring a clean installation. It includes steps to save existing packages, configure npm, update your shell's PATH, and install Claude Code.

```bash
# First, save a list of your existing global packages for later migration
npm list -g --depth=0 > ~/npm-global-packages.txt

# Create a directory for your global packages
mkdir -p ~/.npm-global

# Configure npm to use the new directory path
npm config set prefix ~/.npm-global

# Note: Replace ~/.bashrc with ~/.zshrc, ~/.profile, or other appropriate file for your shell
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc

# Apply the new PATH setting
source ~/.bashrc

# Now reinstall Claude Code in the new location
npm install -g @anthropic-ai/claude-code

# Optional: Reinstall your previous global packages in the new location
# Look at ~/npm-global-packages.txt and install packages you want to keep
# npm install -g package1 package2 package3...
```

--------------------------------

### Instantiate LLM with Advanced Parameters

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

This snippet demonstrates advanced LLM instantiation, including optional parameters for deployment configuration, scaling, hardware, and performance. Adjust parameters like deployment_name, accelerator_type, and replica counts as needed.

```python
llm = LLM(
    model="accounts/fireworks/models/llama-v3p2-3b-instruct",
    deployment_type="on-demand",
    deployment_name="my-custom-deployment",
    accelerator_type="NVIDIA_H100_80GB",
    min_replica_count=1,
    max_replica_count=3,
    scale_up_window=timedelta(seconds=30),
    scale_down_window=timedelta(minutes=10),
    enable_metrics=True
)
```

--------------------------------

### Development Commands

Source: https://github.com/azmaveth/ex_llm/blob/master/CLAUDE.md

Provides essential commands for setting up and running the application during development, including installing dependencies.

```bash
# Run the application
iex -S mix

# Install dependencies
mix deps.get
mix deps.compile
```

--------------------------------

### Use Assistant Prefill for Guided Responses

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

Guide model responses by including a message with `role: "assistant"` at the end of the `messages` array. This is useful for setting a specific tone or starting point for the model's completion.

```typescript
fetch('https://openrouter.ai/api/v1/chat/completions', {
  method: 'POST',
  headers: {
    Authorization: 'Bearer <OPENROUTER_API_KEY>',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'openai/gpt-4o',
    messages: [
      { role: 'user', content: 'What is the meaning of life?' },
      { role: 'assistant', content: "I'm not sure, but my best guess is" },
    ],
  }),
});
```

--------------------------------

### Install Dependencies

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Install the 'datasets' library from Hugging Face to load and prepare data for fine-tuning.

```bash
pip install datasets
```

--------------------------------

### Contextual Prompt Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/perplexity/llms-full.txt

Provide critical context to guide the search, but keep prompts concise and focused on a single topic.

```text
Explain the impact of the 2023 EU digital markets regulations on app store competition for small developers
```

--------------------------------

### Anthropic SDK JavaScript Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/anthropic/llms-full.txt

Demonstrates how to make an API request using the Anthropic SDK in JavaScript. Ensure the SDK is installed and configured.

```javascript
const msg = await client.messages.create({
      model: "claude-3-7-sonnet-v1@20250219",
      max_tokens: 2000,
      temperature: 1,
      system: "Help the user navigate a complex ethical dilemma by identifying core ethical principles, exploring different ethical frameworks, considering potential consequences, acknowledging complexity, encouraging personal reflection, and offering additional resources. Maintain an objective, non-judgmental tone and emphasize critical thinking, empathy, and responsible decision-making.",
      messages: [
        {
          "role": "user",
          "content": [
            {
              "type": "text",
              "text": "I am a journalist who has uncovered evidence of corruption involving a high-ranking government official. Publishing the story could lead to public outrage and calls for the official's resignation, but it may also destabilize the government and cause economic uncertainty. The official has offered me a significant bribe to keep the story quiet. Should I accept the bribe, publish the story, or find another solution?"
            }
          ]
        }
      ]
    });
    console.log(msg);
```

--------------------------------

### List API Keys (PHP)

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

PHP example using GuzzleHttp to retrieve API keys. Ensure GuzzleHttp is installed via Composer.

```php
<?php

$client = new \GuzzleHttp\Client();

$response = $client->request('GET', 'https://openrouter.ai/api/v1/keys', [
  'headers' => [
    'Authorization' => 'Bearer <token>',
  ],
]);

echo $response->getBody();
?>
```

--------------------------------

### Create API Key using Go

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/openrouter/llms-full.txt

This Go program demonstrates creating an API key. Remember to replace '<token>' with your actual token.

```go
package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://openrouter.ai/api/v1/keys"

	payload := strings.NewReader("{\n  \"name\": \"name\"\n}")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Authorization", "Bearer <token>")
	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(res)
	fmt.Println(string(body))

}
```

--------------------------------

### Example Request for Text Completions API

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/lmstudio/llms.txt

Demonstrates how to make a POST request to the /api/v0/completions endpoint to get text completions from a specified model.

```bash
curl http://localhost:1234/api/v0/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "granite-3.0-2b-instruct",
    "prompt": "the meaning of life is",
    "temperature": 0.7,
    "max_tokens": 10,
    "stream": false,
    "stop": "\n"
  }'
```

--------------------------------

### Run Deployment Example Script

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Executes the Python script to test and deploy a reward function to the Fireworks platform. Requires environment variables to be set.

```bash
source .venv/bin/activate && FIREWORKS_API_KEY=$DEV_FIREWORKS_API_KEY FIREWORKS_API_BASE=https://dev.api.fireworks.ai python examples/deploy_example.py
```

--------------------------------

### Refine Search with Latitude and Longitude

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/perplexity/llms-full.txt

Use specific coordinates to get geographically relevant search results. This example approximates San Francisco.

```cURL
curl --location 'https://api.perplexity.ai/chat/completions' \
    --header 'Authorization: Bearer YOUR_API_KEY' \
    --header 'Content-Type: application/json' \
    --data '{ \
      "model": "sonar-pro", \
      "messages": [ \
        {"role": "system", "content": "You are a helpful local guide."}, \
        {"role": "user", "content": "What are some good coffee shops nearby?"} \
      ], \
      "web_search_options": { \
        "user_location": { \
          "latitude": 37.7749, \
          "longitude": -122.4194 \
        } \
      } \
  }'
```

```Python
import requests

url = "https://api.perplexity.ai/chat/completions"
headers = {"Authorization": "Bearer YOUR_API_KEY"}
payload = {
    "model": "sonar-pro",
    "messages": [
        {"role": "system", "content": "You are a helpful local guide."},
        {"role": "user", "content": "What are some good coffee shops nearby?"}
    ],
    "web_search_options": {
      "user_location": {
        "latitude": 37.7749,
        "longitude": -122.4194
      }
    }
}

response = requests.post(url, headers=headers, json=payload)
print(response.json())
```

--------------------------------

### Basic Flight Task Evaluation Example

Source: https://github.com/azmaveth/ex_llm/blob/master/docs/fireworks/llms-full.txt

Example demonstrating how to set the agent model and run evaluation for a basic flight task.

```bash
export MODEL_AGENT=openai/gpt-4o
reward-kit agent-eval --task-dir ./examples/flight_task
```