AmritaSense

"Sense is all you need."

AmritaSense is a general-purpose workflow orchestration engine that replaces traditional graph-based models with an instruction set architecture—treating workflows not as nodes-and-edges diagrams, but as programmable execution streams driven by a lightweight virtual machine.

Why AmritaSense?

Most workflow engines force you into a graph mindset: define nodes, connect edges, manage state objects. AmritaSense takes a different path. You compose nodes and control flow just like writing ordinary code—the engine compiles them into a linear instruction sequence, then executes them step by step. The result: zero scheduling overhead, native interrupt support, and the expressive power of assembly-level control flow.

Core Features

Complete Instruction Set—IF/ELIF/ELSE, WHILE/DO-WHILE, GOTO/CALL, TRY/CATCH/THEN/FIN, NOP, INTERRUPT. All control flow is first-class, not simulated through graph routing.
VM-Style Execution—A program counter (PointerVector) and call stack drive execution. Jumps are integer operations, not graph traversals.
Native REPL Debugger—from amrita_sense.debugger import * gives you breakpoints, stepping, and state inspection. Sync API for interactive REPL use (step, cont); async API (step_async, cont_async) for scripted debugging. Built atop the engine's Panic/Recover mechanism and composite middleware injection.
Async-Native Suspend/Resume—Two Future callbacks enable full workflow interruption at any node boundary. Built for debuggers and human-in-the-loop systems.
Declarative Dependency Injection—Nodes declare dependencies via function signatures. The engine resolves them at runtime with type matching and concurrent resolution.
Ultra Lightweight— Compiles 100,000 nodes in ~200ms. Runs anywhere from Raspberry Pi to cloud.
Self-Compile Instructions—Extend the instruction set with SelfCompileInstruction. Compile-time expansion, zero runtime overhead.

Installation

pip install amrita-sense

Quick Look

import asyncio
from amrita_sense import Node, WorkflowInterpreter as WorkflowPC, IF


@Node()
def condition() -> bool:
    return True


@Node()
def action():
    print("Done")


@Node()
def end():
    print("End of workflow")


flow = IF(condition, action) >> end
pc = WorkflowPC(flow.render())

if __name__ == "__main__":
    asyncio.run(pc.run())

See more demos in demos/

Documentation

Full guides, concept explanations, and API reference at sense.amritabot.com.

Benchmarks

benchmark.py is a self-contained suite covering control flow, jump/call, interrupt, exception, event-dispatch, dependency-injection, composition, concurrency and memory scenarios.

uv run benchmark.py              # 5 measured runs per scenario, prints a table
uv run benchmark.py --runs 20    # more samples, tighter error bars
uv run benchmark.py --list       # list registered scenarios
uv run benchmark.py --no-json    # skip the JSON export

Each scenario is reported as mean ± 1 stdev, and the whole run is written to benchmark-results.json (raw samples plus host metadata), so two revisions can be diffed mechanically instead of eyeballed. CI runs the suite on every supported Python version and uploads the JSON as an artifact.

Contributing

Contributions are welcome. See CONTRIBUTING.md and our Code of Conduct.

License

Apache V2. See LICENSE.

AIGC Content Licensing Policy

AACLP V1. See POLICY_OF_AIGC (Official Link) for details.

Metadata

Release files for amrita-sense 1.1.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for amrita-sense 1.1.2
File	Size	Uploaded
amrita_sense-1.1.2.tar.gz	146.3 kB	Oct 6, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for amrita-sense 1.1.2
File	Interpreter	ABI	Platform	Reset
amrita_sense-1.1.2-py3-none-any.whl 128.9 kB Oct 6, 2026	Python 3	none	any	Details

Total release size: 275.3 kB

Release files / amrita_sense-1.1.2.tar.gz

Download URL	amrita_sense-1.1.2.tar.gz
Size	146.3 kB
Tags	Source
SHA-256 checksum How to use checksums	`e5e528df78848f571d1ba5541b33aa1deda89ac75bb84dc6123c27e47e4a1984`
BLAKE2b-256 checksum How to use checksums	`c777edad80520def333d9327d7ada7f5af366770f1711077e0e365d5fec0c5f1`
Upload date	Oct 6, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`uv/0.12.23 {"installer":{"name":"uv","version":"0.12.23","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}`

Release files / amrita_sense-1.1.2-py3-none-any.whl

Download URL	amrita_sense-1.1.2-py3-none-any.whl
Size	128.9 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`166c211df567f9ab4eb45fc68d68b9580e6a4758819c4666bd22b4688b491ad9`
BLAKE2b-256 checksum How to use checksums	`626e3717ecdd1ea79d37a305ddca615cbc1cffcfb083bf781f59fc4398fc3963`
Upload date	Oct 6, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`uv/0.12.23 {"installer":{"name":"uv","version":"0.12.23","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}`