Pydantic AI Integration¶

Engramia provides EngramiaCapability for Pydantic AI agents — a drop-in capability that recalls patterns before each run and learns from the result.

Installation¶

pip install "engramia[openai,pydantic-ai]"

Quick start¶

from pydantic_ai import Agent
from engramia import Memory
from engramia.providers import OpenAIProvider, OpenAIEmbeddings, JSONStorage
from engramia.sdk.pydantic_ai import EngramiaCapability

mem = Memory(
    llm=OpenAIProvider(model="gpt-4.1"),
    embeddings=OpenAIEmbeddings(),
    storage=JSONStorage(path="./engramia_data"),
)

agent = Agent('openai:gpt-4o', capabilities=[EngramiaCapability(mem)])
result = agent.run_sync("Build a CSV parser")

What happens behind the scenes:

before_run — recalls similar past tasks from Engramia memory
before_model_request — injects recalled patterns into model messages
Agent executes normally
after_run — stores the task + output as a success pattern

EngramiaCapability parameters¶

Parameter	Default	Description
`auto_learn`	`True`	Store successful outputs as patterns
`auto_recall`	`True`	Recall patterns before each run
`min_score`	`7.0`	Eval score assigned when learning
`recall_limit`	`3`	Max patterns to recall per run

System prompt decorator (manual control)¶

For manual control over recall injection:

from pydantic_ai import Agent, RunContext
from engramia.sdk.pydantic_ai import engramia_system_prompt

agent = Agent('openai:gpt-4o')

@agent.system_prompt
def inject_memory(ctx: RunContext) -> str:
    return engramia_system_prompt(mem, ctx, base="You are a Python expert.")

With dependencies¶

from dataclasses import dataclass
from pydantic_ai import Agent

@dataclass
class MyDeps:
    task: str
    user_id: str

agent = Agent(
    'openai:gpt-4o',
    deps_type=MyDeps,
    capabilities=[EngramiaCapability(mem)],
)

result = await agent.run("Build a CSV parser", deps=MyDeps(task="Build a CSV parser", user_id="u1"))

The capability extracts the task from deps.task (or deps.input, deps.query, etc.) for recall.

Structured output¶

from pydantic import BaseModel
from pydantic_ai import Agent

class CodeResult(BaseModel):
    code: str
    explanation: str

agent = Agent('openai:gpt-4o', output_type=CodeResult, capabilities=[EngramiaCapability(mem)])
result = agent.run_sync("Build a CSV parser")
print(result.output.code)
# The full output is learned by Engramia for future recall

Thread safety¶

EngramiaCapability implements for_run() which returns a fresh instance for each run, making it safe for concurrent usage.