Building Multi Agent Automations

Bash
pip install praisonaiagents
export OPENAI_API_KEY=your_key_here

Python
from praisonaiagents import Agent

agent = Agent(instructions="You are a helpful AI assistant")
agent.start("Write a haiku about AI")

For multi-agent systems, use PraisonAIAgents or Workflow. For CLI/no-code usage, install the full praisonai package instead.

Progress checklist for building a PraisonAI automation:

- [ ] 1. Define the task and decide: single agent, multi-agent, or workflow with patterns?
- [ ] 2. Choose LLM provider(s) and set API keys/env vars
- [ ] 3. Create Agent(s) with clear instructions/role/goal
- [ ] 4. Add tools/planning/memory as needed
- [ ] 5. Compose agents into PraisonAIAgents or Workflow
- [ ] 6. Apply workflow patterns (route/parallel/loop/repeat) if orchestration is needed
- [ ] 7. Add safety controls (guardrails, human approval) for production
- [ ] 8. Run, verify output, and tune (memory, context, cost tracking)

Step 1: Decide the architecture

• Single Agent: one task, one instruction set → Agent(...).start(...)
• Multi-Agent (sequential): independent agents handing off results → PraisonAIAgents(agents=[...])
• Workflow: explicit control over step order and branching → Workflow(steps=[...])

Step 2: Create agents

Python
agent = Agent(
    name="Researcher",
    role="Research Analyst",
    goal="Research topics thoroughly",
    instructions="Provide concise, factual information.",
)

Always set instructions; add role/goal for multi-agent clarity.

Step 3: Enhance agents as needed

• Tools: pass Python functions in tools=[...]
• Planning mode: planning=True, planning_tools=[...], planning_reasoning=True — use for multi-step research/tasks requiring upfront decomposition
• Memory: memory=True (zero-dependency file storage) or memory="sqlite" / memory="chromadb" for semantic search; set user_id to isolate per-user state
• Rules/instructions: drop a PRAISON.md, CLAUDE.md, or .cursorrules in project root — auto-discovered and injected into system prompt

Step 4: Compose and orchestrate

Python
from praisonaiagents import Agent, PraisonAIAgents

research_agent = Agent(instructions="Research about AI")
summarise_agent = Agent(instructions="Summarise research agent's findings")

agents = PraisonAIAgents(agents=[research_agent, summarise_agent])
agents.start()

For explicit control, use Workflow with pattern helpers:

Python
from praisonaiagents import Agent, Workflow
from praisonaiagents.workflows import route, parallel, loop, repeat

workflow = Workflow(steps=[
    classifier,
    route({"technical": [tech_agent], "creative": [creative_agent]})
])
result = workflow.start("...")

Step 5: Add safety/observability for production

• Guardrails and human-approval steps for high-risk actions
• Session management / auto-save for long-running workflows
• Telemetry and cost tracking for monitoring

Example 1: Research + Summarize pipeline Input: "Automate researching AI trends and producing a summary." Output:

Python
from praisonaiagents import Agent, PraisonAIAgents

research_agent = Agent(instructions="Research about AI trends in 2025")
summarise_agent = Agent(instructions="Summarise research agent's findings")

agents = PraisonAIAgents(agents=[research_agent, summarise_agent])
agents.start()

Example 2: Query classification and routing Input: "Route technical questions to a tech expert, creative ones to a creative writer." Output:

Python
from praisonaiagents import Agent, Workflow
from praisonaiagents.workflows import route

classifier = Agent(name="Classifier", instructions="Respond with 'technical' or 'creative'")
tech_agent = Agent(name="TechExpert", role="Technical Expert")
creative_agent = Agent(name="Creative", role="Creative Writer")

workflow = Workflow(steps=[
    classifier,
    route({"technical": [tech_agent], "creative": [creative_agent]})
])
result = workflow.start("How do I optimize a Python function for speed?")

Example 3: Agent with persistent memory Input: "Build an assistant that remembers user preferences across sessions." Output:

Python
from praisonaiagents import Agent

agent = Agent(
    name="Personal Assistant",
    instructions="You are a helpful assistant that remembers user preferences.",
    memory=True,
    user_id="user123"
)
agent.start("My name is John and I prefer Python")