Praison AI

Multi-Provider Agent CLI

Command-line interface for running agents with any LLM provider

The CLI runs agents with different LLM providers, tests connectivity, and manages provider configuration.

Diagram

User

CLI

LLM Provider

Response

Quick Start

  1. Simple Usage
    bash
    praisonai-ts llm test openai
praisonai-ts llm run "Hello" --model openai/gpt-4o-mini
  2. With Configuration
    bash
    praisonai-ts llm run "Write a poem" --model openai/gpt-4o-mini --stream
praisonai-ts llm validate openai

Available Commands

CommandDescription
llm providersList available AI SDK providers
llm test <provider>Test connectivity to a provider
llm validate <provider>Validate provider configuration
llm run "<prompt>"Run a prompt with a model
llm modelsList common models by provider
llm configShow resolved configuration
llm traceDemo attribution headers
llm toolsShow tool calling documentation
llm jsonShow structured output documentation

List Providers

List all supported AI SDK providers and their status:

bash
praisonai-ts llm providers

Output:

AI SDK Providers
================
✅ AI SDK: installed

Available Providers:
  ✅ openai               🔑
  ⚠️ anthropic            Missing ANTHROPIC_API_KEY
  ⚠️ google               Missing GOOGLE_API_KEY
  ✅ groq                  🔑
  ...

JSON Output

bash
praisonai-ts llm providers --json
json
{
  "success": true,
  "data": {
    "ai_sdk_available": true,
    "ai_sdk_version": "6.0.3",
    "providers": [
      {
        "id": "openai",
        "package": "@ai-sdk/openai",
        "envKey": "OPENAI_API_KEY",
        "hasApiKey": true,
        "status": "ready"
      }
    ],
    "total": 13
  }
}

Test Provider

Test connectivity to a specific provider:

bash
praisonai-ts llm test openai

Output:

Testing openai/gpt-4o-mini...
✅ Connection successful! (1538ms)
  Provider: openai
  Model: gpt-4o-mini
  Response: OK

With Custom Model

bash
praisonai-ts llm test anthropic --model claude-3-haiku-20240307

JSON Output

bash
praisonai-ts llm test openai --json
json
{
  "success": true,
  "data": {
    "provider": "openai",
    "model": "gpt-4o-mini",
    "status": "success",
    "duration_ms": 1538,
    "response": "OK"
  }
}

Validate Configuration

Validate provider setup including API keys and packages:

bash
praisonai-ts llm validate openai

Output:

Validation: openai
==================
  ✅ AI SDK Installed: AI SDK is installed
  ✅ Provider Supported: Provider 'openai' is supported
  ✅ Provider Package: @ai-sdk/openai is installed
  ✅ API Key: OPENAI_API_KEY is set

✅ All checks passed!

JSON Output

bash
praisonai-ts llm validate anthropic --json
json
{
  "success": true,
  "data": {
    "provider": "anthropic",
    "valid": true,
    "checks": [
      { "name": "AI SDK Installed", "passed": true },
      { "name": "Provider Supported", "passed": true },
      { "name": "Provider Package", "passed": true },
      { "name": "API Key", "passed": true }
    ]
  }
}

Run Prompts

Execute prompts with any supported model:

Basic Usage

bash
praisonai-ts llm run "What is 2+2?"

With Model Selection

bash
praisonai-ts llm run "Explain quantum computing" --model openai/gpt-4o
bash
praisonai-ts llm run "Write a haiku" --model anthropic/claude-3-5-sonnet-latest

Streaming

bash
praisonai-ts llm run "Write a story about a robot" --stream

JSON Output

bash
praisonai-ts llm run "Hello" --json
json
{
  "success": true,
  "data": {
    "model": "openai/gpt-4o-mini",
    "text": "Hello! How can I help you today?",
    "finishReason": "stop",
    "duration_ms": 856
  }
}

With Timeout

bash
praisonai-ts llm run "Complex task" --timeout 120000

Verbose Mode

bash
praisonai-ts llm run "Hello" --verbose

Shows additional details like token usage and timing.

List Models

Show common models for each provider:

bash
praisonai-ts llm models

Output:

Common Models by Provider
=========================
Note: This is not exhaustive. Check provider docs for all models.

  openai:
    - gpt-4o
    - gpt-4o-mini
    - gpt-4-turbo
    - o1
    - o1-mini

  anthropic:
    - claude-3-5-sonnet-latest
    - claude-3-5-haiku-latest
    - claude-3-opus-latest

  google:
    - gemini-2.0-flash
    - gemini-1.5-pro
    - gemini-1.5-flash

Show Configuration

Display resolved configuration with redacted secrets:

bash
praisonai-ts llm config

Output:

AI SDK Configuration
====================
  Defaults:
    timeout: 60000ms
    maxRetries: 2
    maxOutputTokens: 4096
    redactLogs: true

  Provider Aliases:
    oai → openai
    claude → anthropic
    gemini → google

  Environment (redacted):
    OPENAI_API_KEY: ****GW4A
    ANTHROPIC_API_KEY: ****x0QAA

JSON Output

bash
praisonai-ts llm config --json

Attribution Trace Demo

Demonstrate multi-agent attribution headers:

bash
praisonai-ts llm trace

Output:

Attribution Trace Demo
======================
Running with model: openai/gpt-4o-mini

  Attribution Context:
    agentId: agent-mjspa0il
    runId: run-drkw6p18
    traceId: trace-zgu2awc4

  Headers injected:
    X-Agent-Id: agent-mjspa0il
    X-Run-Id: run-drkw6p18
    X-Trace-Id: trace-zgu2awc4

✅ Response: Trace OK (885ms)

Attribution headers are automatically injected into LLM requests.

JSON Output

bash
praisonai-ts llm trace --json
json
{
  "success": true,
  "data": {
    "attribution": {
      "agentId": "agent-mjspa0il",
      "runId": "run-drkw6p18",
      "traceId": "trace-zgu2awc4"
    },
    "headers": {
      "X-Agent-Id": "agent-mjspa0il",
      "X-Run-Id": "run-drkw6p18",
      "X-Trace-Id": "trace-zgu2awc4"
    },
    "response": "Trace OK",
    "duration_ms": 885
  }
}

Tool Calling Documentation

Show how to use tool calling with AI SDK:

bash
praisonai-ts llm tools

Displays example code for defining and using tools with Zod schemas.

Structured Output Documentation

Show how to generate structured JSON output:

bash
praisonai-ts llm json

Displays example code for using generateObject with Zod schemas.

Global Options

All commands support these options:

OptionDescription
--jsonOutput in JSON format
--verboseShow detailed output
--model <provider/model>Specify model
--timeout <ms>Request timeout in milliseconds

Environment Variables

Configure providers via environment variables:

bash
# Required for each provider you want to use
export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GOOGLE_API_KEY=AIza...
export GROQ_API_KEY=gsk_...
export MISTRAL_API_KEY=...
export COHERE_API_KEY=...

Exit Codes

CodeMeaning
0Success
1General error
2Invalid arguments
3Authentication error
4Provider not found
5Network error
6Timeout

Examples

Test All Configured Providers

bash
for provider in openai anthropic google; do
  echo "Testing $provider..."
  praisonai-ts llm test $provider --json
done

Compare Models

bash
PROMPT="Explain AI in one sentence"

echo "OpenAI:"
praisonai-ts llm run "$PROMPT" --model openai/gpt-4o-mini

echo "Anthropic:"
praisonai-ts llm run "$PROMPT" --model anthropic/claude-3-haiku-20240307

echo "Google:"
praisonai-ts llm run "$PROMPT" --model google/gemini-1.5-flash

Streaming with Different Providers

bash
# OpenAI streaming
praisonai-ts llm run "Write a poem" --model openai/gpt-4o-mini --stream

# Anthropic streaming
praisonai-ts llm run "Write a poem" --model anthropic/claude-3-5-sonnet-latest --stream

Troubleshooting

Provider not found
bash
praisonai-ts llm validate <provider>

Install the provider package if needed:

bash
npm install @ai-sdk/<provider>
Authentication error

Ensure your API key is set:

bash
export OPENAI_API_KEY=sk-...
praisonai-ts llm test openai
Timeout issues

Increase timeout for slow requests:

bash
praisonai-ts llm run "Complex task" --timeout 120000

Related

AI SDK

Code-based AI SDK usage

 Streaming

Advanced streaming

 Provider Registry

Managing providers