Skip to main content

Modular Reasoning Framework: Multi-stage constraint layer for autonomous agents

Project description

Modular Reasoning Framework (MRF)

MRF is a lightweight, transparent, multi-stage reasoning scaffold for autonomous agents. It provides a decomposable structure built from reusable computational primitives (Transform, Evaluate, Filter, Summarize, Reflect, Inspect, Rewrite) to make agent behavior traceable and reliable.


Why MRF

Autonomous agents fail because:

  1. No reasoning structure
  2. No constraint layer
  3. No diagnostics

MRF introduces explicit stages, bounded operations, and full visibility into how reasoning unfolds.


Features

• Reusable Operators

Transform
Evaluate
Filter
Summarize
Reflect
Inspect
Rewrite

• Multi-Stage Pipeline

Understanding → Planning → Execution → Synthesis → Verification

• Diagnostics

Stage logs
Confidence tracking
Goal alignment
Failure detection

• Security-Relevant Mitigation

MRF is not an alignment system, but adds structure that reduces unbounded behavior, prompt-injection surfaces, and silent drift.


Install

pip install mrf-core

Quick Start

from mrfcore.pipeline import ReasoningPipeline

pipeline = ReasoningPipeline(verbose=True) result = pipeline.run("Plan a three-step morning routine.")

print(result["answer"]) print(result["confidence"]) print(result["valid"])


Inspecting a Stage

pipeline.inspect_stage(result, "understanding")


Custom Pipeline

from mrfcore.phases import UnderstandingPhase, PlanningPhase, ExecutionPhase from mrfcore.pipeline import Pipeline

pipeline = Pipeline([ UnderstandingPhase(), PlanningPhase(), ExecutionPhase() ])

result = pipeline.run("Help me clean my desk efficiently.")


Project Structure

mrfcore/ operators/ base.py transform.py evaluate.py filter.py summarize.py reflect.py inspect.py rewrite.py engine.py pipeline.py phases.py presets.py diagnostics.py exceptions.py examples/ tests/


What MRF Is Not

Not alignment Not a safety guarantee Not a sandbox Not a replacement for secure execution layers MRF adds structure


Contributing

PRs welcome


License

License

License: Apache 2.0

This project is licensed under the Apache License 2.0.

Copyright 2026 Ryan Sabouhi

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License.

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

You may obtain a copy of the License at: https://www.apache.org/licenses/LICENSE-2.0


Notice

This software contains original work developed as part of the
Modular Reasoning Framework (MRF) by Ryan Sabouhi.
See the NOTICE file for additional attribution details.

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

mrf_core-1.0.0.tar.gz (18.7 kB view details)

Uploaded Source

Built Distribution

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

mrf_core-1.0.0-py3-none-any.whl (18.7 kB view details)

Uploaded Python 3

File details

Details for the file mrf_core-1.0.0.tar.gz.

File metadata

  • Download URL: mrf_core-1.0.0.tar.gz
  • Upload date:
  • Size: 18.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.5

File hashes

Hashes for mrf_core-1.0.0.tar.gz
Algorithm Hash digest
SHA256 aa6f8d15c598c1c8515c83d385f5693f16b4ec18f7a8324cfea35f474f2d37f2
MD5 df015f018ebf4e2a3a7ae570d556f01b
BLAKE2b-256 d33d835882823095356c938a55af87980dc2a33aa49efa8ce3b11b850fd82460

See more details on using hashes here.

File details

Details for the file mrf_core-1.0.0-py3-none-any.whl.

File metadata

  • Download URL: mrf_core-1.0.0-py3-none-any.whl
  • Upload date:
  • Size: 18.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.5

File hashes

Hashes for mrf_core-1.0.0-py3-none-any.whl
Algorithm Hash digest
SHA256 c893cf6e251470e64e120e3a9be12d5a655caa8857f4dd683f4b5f233cce7595
MD5 c6d652105ca95eaffb9d75aa0af2f1dd
BLAKE2b-256 13894dee37422ff9cf72a12a0ad46f7a4d124e559947347a11f80bec58a6c257

See more details on using hashes here.

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