Capabilities Overview

LiteLLM endpoint parity capabilities for PraisonAI

Next

Overview

PraisonAI Capabilities provide direct access to LiteLLM endpoints with full parity, enabling you to use completions, embeddings, images, audio, and more through a unified API.

Available Capabilities

Core APIs

Media Generation

Safety & Moderation

Advanced

Quick Start

from praisonai import embed, embedding  # Both work identically
from praisonai.capabilities import chat_completion, image_generate, moderate

# Chat completion
result = chat_completion(
    messages=[{"role": "user", "content": "Hello!"}],
    model="gpt-4o-mini"
)
print(result.content)

# Embeddings (embed and embedding are aliases)
result = embed("Hello world", model="text-embedding-3-small")
print(f"Dimensions: {len(result.embeddings[0])}")

# Image generation
result = image_generate("A sunset", model="dall-e-3")
print(f"URL: {result[0].url}")

# Moderation
result = moderate("Check this content")
print(f"Flagged: {result[0].flagged}")

CLI Quick Start

# Chat completion
praisonai completions "What is AI?"

# Embeddings (both commands work)
praisonai embed "Hello world"
praisonai embedding "Hello world"

# Image generation
praisonai images "A beautiful sunset"

# Moderation
praisonai moderate "Check this content"

All Capabilities

The full list of 116 capabilities includes:

• Audio: transcribe, speech
• Images: image_generate, image_edit
• Videos: video_generate (default openai/sora-2)
• Files: file_create, file_list, file_retrieve, file_delete
• Batches: batch_create, batch_list, batch_retrieve, batch_cancel
• Vector Stores: vector_store_create, vector_store_search, vector_store_file_create, vector_store_file_list, vector_store_file_delete
• Embeddings: embed, embedding (alias)
• Rerank: rerank
• Moderations: moderate
• OCR: ocr
• Assistants: assistant_create, assistant_list
• Fine-tuning: fine_tuning_create, fine_tuning_list
• Responses: responses_create
• Passthrough: passthrough (api_base is DNS-resolved and screened; localhost and private/metadata addresses are rejected, see Passthrough SSRF Guard)
• Containers: container_create (no backend yet — raises NotImplementedError; not in praisonai.capabilities.__all__, see Containers)
• Search: search (requires the ddgs package; raises on error, see Search)
• A2A: a2a_send
• Completions: chat_completion, text_completion
• Messages: messages_create, count_tokens
• Guardrails: apply_guardrail — fails closed by default; on_error="allow" for the old permissive behaviour
• RAG: rag_query
• Realtime: realtime_connect
• Skills: skill_list, skill_load
• MCP: mcp_list_tools, mcp_call_tool

All capabilities have both sync and async versions (prefixed with a).