Examples

All examples live in examples/.

Setup

# Copy and fill in your API keys
cp .env.example .env

# Run any example
cargo run --example <name>

---

OpenAI

examples/openai.rs — Chat · Streaming · Responses API · Embeddings

cargo run --example openai

Covers five scenarios in one shot:

1. Basic chat (gpt-4o)
2. Streaming (with_stream(true)) with real-time token output
3. Responses API (invoke_responses)
4. Responses streaming (invoke_responses_stream)
5. Embeddings (text-embedding-3-small) — returns vector size

OPENAI_API_KEY=sk-...
OPENAI_BASE_URL=https://api.openai.com/v1   # optional override

---

Anthropic Claude

examples/anthropic.rs — System messages · Streaming

cargo run --example anthropic

1. Chat with a system message (persona: speaks like a pirate)
2. Streaming response for a long-form question

ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_BASE_URL=https://api.anthropic.com   # optional

---

Google Gemini

examples/google.rs — Chat · Streaming · Native Thinking

cargo run --example google

1. Basic chat with Gemini 1.5 Flash
2. Streaming chat
3. Reasoning / thinking mode (with_enable_thinking(true)) — exposes separate reasoning_content

examples/google_tools_thinking.rs — Tools + Thinking combined

cargo run --example google_tools_thinking

Full 3-step flow: tool call request → mock tool execution → tool result → final answer, all with enable_thinking active. Uses Gemini 3 Flash Preview / 2.0.

GOOGLE_API_KEY=...
GOOGLE_REGION=global   # or cn
GOOGLE_MODEL=gemini-3-flash-preview

examples/list_google_models.rs — List available models

cargo run --example list_google_models

---

Aliyun DashScope (Qwen)

examples/aliyun.rs — Chat · Streaming · Reasoning

cargo run --example aliyun

1. Basic chat (qwen-max)
2. Streaming chat
3. Reasoning / thinking mode (qwen-plus + with_enable_thinking(true))

ALIYUN_API_KEY=sk-...
ALIYUN_BASE_URL=https://dashscope.aliyuncs.com   # optional

---

Zhipu GLM

examples/zhipu.rs — Multi-region · Chat · Streaming

cargo run --example zhipu

Supports domestic (cn / bigmodel.cn) and global (global / z.ai) endpoints via ZHIPU_REGION. Optional proxy via ZHIPU_PROXY.

ZHIPU_API_KEY=...
ZHIPU_REGION=cn           # or global
ZHIPU_MODEL=glm-4.5-flash
ZHIPU_PROXY=http://127.0.0.1:7890   # optional

examples/zhipu_tools.rs — Tool Calling (GLM-5, Global)

cargo run --example zhipu_tools

Full 2-step tool call flow with reasoning output before the tool call decision.

examples/zhipu_thinking.rs — Reasoning / Thinking (GLM-5)

cargo run --example zhipu_thinking

Uses with_enable_thinking(true) for step-by-step maths reasoning; displays reasoning_content separately from the final answer.

examples/zhipu_vision.rs — Image Understanding (GLM-4V)

cargo run --example zhipu_vision

Sends a URL image + text question using MessageBlock::image_url(...). Works with glm-4v and glm-5.

---

Moonshot (Kimi)

examples/moonshot.rs — Multi-region · Chat · Streaming

cargo run --example moonshot

Selects endpoint via MOONSHOT_REGION (cn → api.moonshot.cn, global → api.moonshot.ai).

MOONSHOT_API_KEY=sk-...
MOONSHOT_REGION=cn
MOONSHOT_MODEL=kimi-k2.5

examples/moonshot_tools.rs — Tool Calling

cargo run --example moonshot_tools

2-step weather tool flow: request tool → parse typed args → return result → final answer.

examples/moonshot_thinking.rs — Reasoning / Thinking

cargo run --example moonshot_thinking

Logic puzzle solved with with_enable_thinking(true); reasoning_any() extracts the chain-of-thought across multiple field spellings (reasoning_content, reasoning, thought, thinking).

---

DeepSeek

examples/deepseek.rs — Basic Chat

cargo run --example deepseek

Minimal example using LlmClient::openai(key, base_url) pointed at api.deepseek.com. DeepSeek uses the OpenAI wire format.

DEEPSEEK_API_KEY=sk-...
DEEPSEEK_BASE_URL=https://api.deepseek.com   # optional

---

Minimax

examples/minimax.rs — Multi-region · Chat · Streaming

cargo run --example minimax

Uses openai_compatible builder, selects global endpoint by default. Minimax M2.5 via OpenAI-compatible API.

MINIMAX_API_KEY=...
MINIMAX_REGION=global
MINIMAX_MODEL=MiniMax-M2.5

---

Tencent Hunyuan

examples/tencent.rs — Chat · Streaming (TC3-HMAC-SHA256)

cargo run --example tencent --features tencent

Requires the tencent Cargo feature. Native Tencent Cloud API v3 signing (SecretId + SecretKey).

TENCENT_SECRET_ID=AKID...
TENCENT_SECRET_KEY=...
TENCENT_BASE_URL=hunyuan.tencentcloudapi.com

---

Ollama (Local Models)

examples/ollama.rs — Model listing · Chat · Streaming

cargo run --example ollama

No API key required. Connects to http://localhost:11434 by default.

1. List locally pulled models
2. Chat with llama3
3. Streaming chat

Make sure Ollama is running and the model is pulled:

ollama pull llama3

---

Cross-Provider Examples

examples/tool_calling.rs — Full 2-step Tool Call (OpenAI)

cargo run --example tool_calling

The reference tool-calling example. Defines a get_current_weather function, sends it via OpenAI gpt-4o-mini, executes the mock tool, and returns the result for a final answer.

examples/multi_modal.rs — Image Analysis + Embeddings (OpenAI)

cargo run --example multi_modal

Combines multi-modal streaming (image URL → MessageBlock::image_url(...)) with embeddings in a single run.

examples/check_reasoning_formats.rs — Reasoning Field Inspection

cargo run --example check_reasoning_formats

Inspects all reasoning-related field variants (reasoning_content, reasoning, thought, thinking) across multiple providers to verify extraction.

examples/real_world_connectivity_test.rs — Batch Connectivity Test

cargo run --example real_world_connectivity_test

Tests every configured provider in sequence and prints a pass/fail summary. Useful as a smoke test before deployment.

---

Environment Variables Reference

Provider	Variable(s)
OpenAI	OPENAI_API_KEY, OPENAI_BASE_URL
Anthropic	ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL
Google	GOOGLE_API_KEY, GOOGLE_REGION, GOOGLE_MODEL, GOOGLE_BASE_URL
Aliyun	ALIYUN_API_KEY, ALIYUN_BASE_URL
Zhipu	ZHIPU_API_KEY, ZHIPU_REGION, ZHIPU_MODEL, ZHIPU_PROXY
Moonshot	MOONSHOT_API_KEY, MOONSHOT_REGION, MOONSHOT_MODEL
DeepSeek	DEEPSEEK_API_KEY, DEEPSEEK_BASE_URL
Minimax	MINIMAX_API_KEY, MINIMAX_REGION, MINIMAX_MODEL
Tencent	TENCENT_SECRET_ID, TENCENT_SECRET_KEY, TENCENT_BASE_URL
Ollama	(none — defaults to localhost:11434)