pyagent-all 0.2.3

Meta-package: installs all PyAgent packages (patterns, router, compress, trace, providers, context, blueprint, studio)

pyagent-all

Meta-package that installs the complete PyAgent suite. One command, everything included.

Install

pip install pyagent-all

What's Included

Package	Description
pyagent-patterns	18 multi-agent orchestration patterns + composites + guardrails + recovery
pyagent-router	Difficulty scoring, cost estimation, model selection, routing middleware
pyagent-compress	Inter-agent message compression, agent pruning, interaction pruning, token budgets
pyagent-trace	TraceEventBus pub/sub, OpenTelemetry spans, Langfuse export, cost tracking, record/replay
pyagent-providers	Multi-provider abstraction, registry, routing strategies, fallback chains, capability negotiation, cost optimizer
pyagent-context	Structured context with trust/sensitivity metadata, three-tier memory (working/session/semantic), compression, retrieval, redaction
pyagent-blueprint	Declarative YAML specs, Pydantic validation, compilation to RuntimeGraph, contract testing, Mermaid rendering, semantic diff, CLI
pyagent-studio	CLI + web control plane for designing, simulating, debugging, and governing agent blueprints with live trace streaming

Quick Start

import asyncio
from pyagent_patterns.base import Agent, MockLLM
from pyagent_patterns.orchestration import Pipeline
from pyagent_router import ModelSelector
from pyagent_compress import MessageCompressor, CompressMiddleware, TokenBudget
from pyagent_trace import CostTracker
from pyagent_trace.events import TraceEventBus
from pyagent_providers import ProviderRegistry, MockProvider
from pyagent_context import ContextLedger, ContextItem, TrustLevel

# Set up tracing
bus = TraceEventBus()
tracker = CostTracker(event_bus=bus)

# Set up compression with token budget
budget = TokenBudget(workflow_limit=50_000, per_agent_limit=10_000)
middleware = CompressMiddleware(
    compressor=MessageCompressor(target_ratio=0.5),
    budget=budget,
)

# Build and run a pipeline
llm = MockLLM(responses=["Key facts extracted", "Concise summary"])
pipeline = Pipeline(stages=[
    middleware.wrap(Agent("extractor", llm, system_prompt="Extract key facts.")),
    Agent("summarizer", llm, system_prompt="Write a concise summary."),
])

result = asyncio.run(pipeline.run("Process this document"))
print(result.output)         # "Concise summary"
print(budget.summary())      # Token usage per agent
print(tracker.summary())     # Cost breakdown by pattern/agent/model

End-to-End Flow

The packages are designed to work together in a layered architecture:

1. Blueprint → Define agents, workflows, providers, contracts, observability, and context in YAML
2. Compile → BlueprintCompiler produces a RuntimeGraph of patterns and agents
3. Orchestrate → Patterns (Pipeline, Supervisor, Debate, etc.) coordinate agents
4. Provide → ProviderRegistry + FallbackChain handle LLM calls with resilience
5. Trace → TraceEventBus collects events from agents, patterns, providers, and compression
6. Compress → CompressMiddleware reduces inter-agent token transfer; TokenBudget enforces limits
7. Remember → ContextLedger with three-tier memory (working → session → semantic) persists context
8. Observe → Studio dashboard visualizes traces, costs, governance, and agent communication

Documentation

See pyagent.org for full docs, API reference, and cookbook.