Skip to main content

Deterministic Context Assembly for AI Agents.

Project description

Agent Assembler 🧩

The On-Demand Digital Arsenal for the AI Era. Deterministic Context Assembly for Multi-Agent Distribution Network.

v0.5.0 Released PyPI

We don't build the Aircraft Carriers (LLMs); we build the 4S Shop. Agent Assembler is the core engine that powers multi-agent systems, turning non-standard business requirements into standardized, executable AI Agents.


🚀 Why Agent Assembler?

In the current AI landscape, Large Language Models provide immense "intelligence," yet they often fail in critical business scenarios due to a lack of Domain Knowledge and Deterministic Logic.

Feature Traditional SaaS Traditional Outsourcing Agent Assembler
Delivery Rigid, standard accounts Expensive, slow human labor On-demand digital production
Flexibility Low (Roadmap dependent) High (Cost scales linearly) Extreme (Sidecar hot-plugging)
Core Asset Vendor's platform code Client's private code The Recipe Matrix (Industry Wisdom)
Marginal Cost Low (per seat) High (per project) Near Zero (per assembly)

Agent Assembler bridges the gap with a "Recipe + Sidecar" architecture, allowing domain experts to convert tacit industry knowledge into executable digital assets.


🏗️ Core Architecture

1. Micro-kernel + Sidecar Bus

The core soul of the framework — decoupling core logic from auxiliary capabilities for maximum flexibility.

  • 🥘 Pot (The Recipe): Encapsulates essential business logic and data flow (e.g., Profit Calculation, Compliance Check).
  • 🔌 Sidecar Bus (Pluggable Plugins): Capabilities are hot-swappable via a standard bus:
    • 🧠 Decision Engine: Deterministic analysis & "Red-Yellow-Green" verdicts.
    • 🎭 Simulator: Immersive role-play for business negotiation training.
    • 📊 Analytics: Data persistence and visualization pipelines.
  • 🏭 Factory (The Assembler): Orchestrates the "Pots" and "Lids" to assemble and deploy specific Agents.

2. JIT Assembly & Atomic Design

  • Recipe-First: Intent matching against pre-defined JSON recipes.
  • Atomic Skills (<4KB): Focused, composable modules that do one thing well.
  • AutoCraft: Automated generation engine transforming unstructured requirements into deployed scripts with zero-marginal-cost.

3. Multi-Platform Adapters

Deploy assembled Agents to any platform with one click: Qianwen, Coze, WeChat, and more.


🛠️ Installation & Quick Start

pip install agent-assembler
from agent_assembler import Assembler, Agent, AgentSpec, SidecarBus

# Quick JIT assembly
assembler = Assembler(recipes_dir="./recipes", skills_dir="./skills")
result = assembler.assemble("Analyze this excel file")
print(result['system_prompt'])

# Spec-driven Agent assembly
spec = AgentSpec(name="Tax Advisor", role="财务顾问", recipes=["tax_consulting"])
agent = assembler.assemble_agent(spec)

# Sidecar Bus (pluggable capabilities)
bus = SidecarBus(decision_engine=True, simulator=True, analytics=True)

🎯 From Sandbox to Matrix

Our architecture is Domain-Agnostic. It works anywhere.

  1. 🧪 The Sandbox (Validation Phase)

    • Validated in high-noise, non-standard environments (e.g., Tier-3/4 city retail & dining markets).
    • Proven: Solves "messy accounts," compliance risks, and performance disputes with "Micro-Deductions" (calculating exact break-even points in seconds).
  2. 🌍 The Matrix (Expansion Phase)

    • Zero Code Change required at the core engine level.
    • Cross-Border E-commerce: Inventory turnover optimization.
    • Manufacturing: Production line yield simulation.
    • Compliance: Automated contract auditing.

🗺️ Roadmap

Phase Goal Status
P0 Core Stabilization & Validation ✅ Done
P1 Architecture Refactoring (api_gateway → modular) ✅ Done
P2 Bug Fixes (schema, paths, resource leaks, test coverage) ✅ Done
P3 Documentation Alignment ✅ Done
P4 SDK Hardening (Agent/AgentSpec/Sidecar Bus) ✅ Done
P4.3 Recipe Registry (search, tags, version management) ✅ Done
P4.4 LLM Full Chain (LLMClient, Agent.run) ✅ Done
Sidecar Decision Engine, Simulator, Analytics ✅ Done
P5 SaaS Dashboard & No-Code Builder ✅ Done
AutoCraft v2 Schema fix, .md skills, validation ✅ Done

🔮 Vision

We are building the Salesforce of the AI Era.

We provide not just tools, but an infrastructure for global developers and industry experts to co-build a Commercial Wisdom Recipe Library.

"The future is already here — it's just not evenly distributed. We are the pipeline builders."


📜 License

Apache 2.0 (Core SDK) / Commercial (Proprietary Recipes)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

agent_assembler-0.5.0.tar.gz (39.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

agent_assembler-0.5.0-py3-none-any.whl (32.7 kB view details)

Uploaded Python 3

File details

Details for the file agent_assembler-0.5.0.tar.gz.

File metadata

  • Download URL: agent_assembler-0.5.0.tar.gz
  • Upload date:
  • Size: 39.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for agent_assembler-0.5.0.tar.gz
Algorithm Hash digest
SHA256 13d3d7cfeef20014b0323cfb29cf58b16cb79c79146b20aa295ad7073443d648
MD5 c5cce246dbe3b39b08fd75b218d166d7
BLAKE2b-256 f57cf42000c81eef2b837554156ca9418d757a904c6887905aa795b994e5725e

See more details on using hashes here.

Provenance

The following attestation bundles were made for agent_assembler-0.5.0.tar.gz:

Publisher: publish.yml on maming0203/agent-assembler

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file agent_assembler-0.5.0-py3-none-any.whl.

File metadata

File hashes

Hashes for agent_assembler-0.5.0-py3-none-any.whl
Algorithm Hash digest
SHA256 044cea7a0d122b8ccef2684519b3b0dcc87cbdfb3ae3df837380da41983a4351
MD5 6d6dd0b7f16a7535e4681b7b6c442529
BLAKE2b-256 e5c987f8cb69e42ce6bfbb1a5601a933883b3e5ce27e32ee8d7a345209704775

See more details on using hashes here.

Provenance

The following attestation bundles were made for agent_assembler-0.5.0-py3-none-any.whl:

Publisher: publish.yml on maming0203/agent-assembler

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page