bifrost-ai 1.0.0

Unified Python SDK for multiple AI providers with automatic conversation management

BifrostAI Python SDK

Official Python SDK for BifrostAPI — one API for various AI models (ChatGPT, Claude, Gemini).

Get Your API Key

Subscribe to BifrostAPI on one of these platforms:

• RapidAPI

Installation

pip install bifrost-ai

Quick Start

from bifrost import BifrostAI

client = BifrostAI(
    api_key="bfr-xxxxx",
    openai_key="sk-xxxxx",
    storage="memory"
)

response = client.chat("conv1", "Hello!", model="gpt-4")
print(response.content)

Storage Options

The storage parameter is required when using methods with conversation_id (automatic history management).

Method	Storage Required
client.chat()	✅ Yes
client.chat_with_files()	✅ Yes
client.chat_stream()	✅ Yes
client.create_chat_completion()	❌ No

Storage backends:

Type	Format	Persistence
Memory	"memory"	❌ Lost on restart
File	"file://path.json"	✅ Saved to disk
Redis	"redis://localhost:6379"	✅ Saved to Redis
PostgreSQL	"postgresql://user:pass@host/db"	✅ Saved to DB

High-level API (requires storage)

client = BifrostAI(api_key="...", openai_key="...", storage="memory")

# SDK manages conversation history automatically
response = client.chat("conv1", "My name is Farid", model="gpt-4")
response = client.chat("conv1", "What is my name?", model="gpt-4")  # Remembers!

Low-level API (no storage needed)

client = BifrostAI(api_key="...", openai_key="...")

# You manage messages yourself
response = client.create_chat_completion(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello"}]
)

Streaming

for chunk in client.chat_stream("conv1", "Count 1 to 5", model="gpt-4"):
    print(chunk.content, end="", flush=True)

File Upload

# Simple - just pass file paths
response = client.chat_with_files(
    "conv1",
    "Summarize this document",
    files=["document.pdf"],  # Same folder
    model="gpt-4-turbo"
)
print(response.content)

# Full path
response = client.chat_with_files(
    "conv1",
    "What is in this file?",
    files=[r"C:\Users\You\Documents\report.pdf"],
    model="gpt-4-turbo"
)

# Multiple files
response = client.chat_with_files(
    "conv1",
    "Compare these documents",
    files=["doc1.pdf", "doc2.pdf"],
    model="gpt-4-turbo"
)

License

MIT