100 Libraries. Zero Dependencies. Pure Dharma.

कर्मण्येवाधिकारस्ते मा फलेषु कदाचन

"You have the right to work, but never to the fruit of work." — Bhagavad Gita 2.47

धर्मDharma
विचारVichara
कर्मKarma
सेवाSeva
लोकLoka
100Libraries
4,375Tests
0Ext Dependencies
herald
Semantic routing engine. Routes queries to the right agent using vector similarity and intent classification.
48 tests GitHub
from herald import Router
router = Router()
result = router.route("fix this bug")
engram
5-tier memory system with semantic recall, boosting, and compaction. Qdrant vectors + Redis hot cache.
62 tests GitHub
from engram import Memory
mem = Memory(tiers=5)
mem.store("user prefers dark mode")
sentinel
Input/output guard system. Validates, sanitizes, and shields agent pipelines from prompt injection.
55 tests GitHub
from sentinel import Guard
guard = Guard(strict=True)
safe = guard.check(user_input)
verdict
LLM evaluation framework. Judges response quality, hallucination detection, and coherence scoring.
44 tests GitHub
from verdict import Judge
score = Judge().evaluate(
  response, ground_truth)
karma
Action execution engine. Converts agent decisions into tool calls, API requests, and file operations.
38 tests GitHub
from karma import Executor
ex = Executor(retry=3)
ex.run("deploy_service", env="prod")
dharma
Ethics and alignment framework. Enforces constitutional AI principles and harm prevention.
36 tests GitHub
from dharma import Ethics
ethics = Ethics(constitution="v2")
ethics.validate(agent_action)
sutra
Configuration threading. Weaves config across distributed agents with hot-reload and schema validation.
31 tests GitHub
from sutra import Config
cfg = Config.load("agents.toml")
cfg.watch(on_change=reload)
shakti
Power management for agent pools. Scales workers up/down based on load, manages GPU allocation.
29 tests GitHub
from shakti import Pool
pool = Pool(min=2, max=20)
pool.autoscale(metric="queue_depth")
maya
Illusion testing framework. Generates adversarial inputs, edge cases, and synthetic data for agent testing.
42 tests GitHub
from maya import Adversarial
tests = Adversarial.generate(
  prompt, n=100)
brahma
Agent creation toolkit. Scaffolds new agents with memory, tools, and guardrails in a single call.
35 tests GitHub
from brahma import create_agent
agent = create_agent(
  name="coder", model="codestral")
mohini
The Agent OS. Orchestrates all 20 specialist agents, manages lifecycle and inter-agent communication.
72 tests GitHub
from mohini import Orchestrator
orch = Orchestrator(agents=20)
orch.run_forever()
drashta
Observability toolkit. Traces agent execution, logs decisions, measures latency in real time.
41 tests GitHub
from drashta import Tracer
tracer = Tracer(service="herald")
tracer.span("route_query")

+ 88 more libraries in the arsenal. Each battle-tested, zero external dependencies.

Why Zero Dependencies?

Every library in the arsenal ships with zero external dependencies. This is not a constraint. It is a philosophy.

Supply Chain SecurityNo transitive vulnerabilities. No left-pad incidents. Every line of code is audited and owned.
Deterministic BuildsSame source, same binary, every time. No version resolution surprises across environments.
Deep UnderstandingWhen you write every component, you understand every failure mode. Debugging is instant.
Minimal FootprintSmaller binaries. Faster cold starts. Lower memory usage. Better for edge and serverless.
कर्मण्येवाधिकारस्ते मा फलेषु कदाचन
You have the right to work, but never to the fruit of work. — Bhagavad Gita 2.47