aimmh-lib 1.1.0

Async multi-model conversation orchestration (fan_out, daisy_chain, room_all, room_synthesized, council, roleplay, MultiModelHub, ModelInstance)

Emergent — Multi-Model AI Hub

Chat with GPT, Claude, Gemini, Grok, DeepSeek, and Perplexity in parallel. Compare, synthesize, and chain their responses — all in one interface.

---

Try it live

emergentapp.interdependentway.org

Tier	Price	What you get
Free	$0	5 instances, 10 runs/month, basic responses and chat
Supporter	$5/month	15 instances, 30 runs/month, removes "Made with Emergent" badge, Hall of Makers eligibility, early feature voting, private thank-you channel
Coffee	$5 one-time	Supporter perks + Hall of Makers eligibility
Builder	$25 one-time	Supporter perks + Hall of Makers eligibility (Builder tier)
Patron	$50 one-time	Supporter perks + Hall of Makers eligibility (Patron tier)
Pro	$19/month or $149/year	Unlimited instances & runs, advanced synthesis, priority support, all supporter perks
Team	$49/month (3 seats)	Unlimited instances & runs, shared workspace, admin controls, all supporter perks
Team extra seat	$15/month	Adds 1 extra seat to an existing Team plan

---

What is this?

Emergent is a multi-model AI hub that lets you send one prompt to multiple LLMs simultaneously and work with their responses together. Instead of copy-pasting between chat tabs, you get a single interface with color-coded, side-by-side or stacked responses from every model you care about.

Beyond simple fan-out, Emergent supports structured interaction patterns — synthesis (feed multiple responses into one model for analysis), shared rooms (models that see and respond to each other), daisy chains (A→B→C sequential pipelines), council mode, and roleplay scenarios. The EDCM engine analyzes conversation transcripts across six cognitive metrics and surfaces actionable insights.

---

Interaction Patterns

Pattern	What it does
Fan-out	Send one prompt to N models in parallel
Synthesis	Select responses, send to a synthesis model for analysis
Shared Room (All)	All models see each other's responses and reply in rounds
Shared Room (Synthesized)	Responses synthesized first, then drive the next round
Daisy Chain	Model A → B → C sequentially, each seeing the previous response
Council	Each model synthesizes all responses including its own
Roleplay	DM-driven roleplay with initiative ordering and reactions

---

Self-hosting

Backend (FastAPI + MongoDB)

cd backend
pip install -r requirements.txt

# Required env vars
export MONGO_URI="mongodb://localhost:27017"
export JWT_SECRET="your-secret"
export STRIPE_SECRET_KEY="sk_..."        # optional: for payments
export STRIPE_WEBHOOK_SECRET="whsec_..." # optional: for webhooks

uvicorn server:app --reload

Frontend (React)

cd frontend
npm install
npm start

The frontend expects the backend at http://localhost:8000 by default.

---

aimmh-lib — the open-source core

The orchestration patterns are extracted into a standalone, zero-dependency Python library.

pip install aimmh-lib

Functional API

import asyncio
from aimmh_lib import fan_out

async def call_model(model_id: str, messages: list[dict]) -> str:
    # plug in any model backend here
    return f"Response from {model_id}"

async def main():
    results = await fan_out(
        call_fn=call_model,
        model_ids=["gpt-4o", "claude-sonnet-4-6", "gemini-2.0-flash"],
        messages=[{"role": "user", "content": "What is the best programming language?"}],
    )
    for r in results:
        print(f"{r.model_id}: {r.content}")

asyncio.run(main())

Instantiation API

Bind a backend once and call any pattern as a method:

from aimmh_lib import MultiModelHub

hub = MultiModelHub(call_model)

results = await hub.fan_out(["gpt-4o", "claude-sonnet-4-6"], messages)
results = await hub.daisy_chain(["gpt-4o", "claude-sonnet-4-6", "gemini-2.0-flash"], "Explain gravity")
results = await hub.council(["gpt-4o", "claude-sonnet-4-6"], "What is consciousness?")

Stateful single-model conversations

from aimmh_lib import ModelInstance

gpt = ModelInstance(call_model, "gpt-4o", system_context="You are a Socratic tutor.")
r1 = await gpt.send("What is entropy?")
r2 = await gpt.send("Give me an example.")  # history carries over automatically
print(gpt.history)
gpt.clear()  # reset to fresh state

All six patterns available: fan_out, daisy_chain, room_all, room_synthesized, council, roleplay.

PyPI →

---

Tech Stack

Backend: FastAPI · Motor (async MongoDB) · asyncio · Stripe · Google OAuth · JWT

Frontend: React · Tailwind CSS · Shadcn UI · React Router

Library: Pure Python 3.11+ · zero runtime dependencies

---

Repository Structure

aimmh_lib/   # pip install aimmh-lib — zero-dep async orchestration library
backend/     # FastAPI service (auth, multi-model chat, payments, EDCM)
frontend/    # React UI

---

License

aimmh_lib/ is licensed under the Apache License 2.0. The backend and frontend are proprietary — you may self-host for personal use but may not offer them as a competing hosted service.