agenticmem 0.1.3.7

A Python client for the AgenticMem API

AgenticMem Python Client

A Python client library for interacting with the AgenticMem API. This client provides easy-to-use interfaces for managing user interactions, profiles, configuration, and agent feedback.

Installation

pip install agenticmem

Quick Start

import agenticmem
from agenticmem import (
    InteractionRequest,
    SearchUserProfileRequest,
    SearchInteractionRequest,
    UserActionType
)
from datetime import datetime

# Initialize the client
client = agenticmem.AgenticMemClient(url_endpoint="http://127.0.0.1:8081/")

# Login with email/password (async)
token = await client.login("[e-mail]", "password")
client.api_key = token.api_key

# Publish a user interaction (async)
response = await client.publish_interaction(
    request_id="your_request_id",
    user_id="your_user_id",
    interaction_requests=[
        InteractionRequest(
            role="User",
            content="Hey, this is John. How can I help you today?",
            user_action=UserActionType.NONE,
            user_action_description="",
            interacted_image_url=""
        )
    ],
    source="conversation"
)
print(f"Published interaction: {response.success} - {response.message}")

# Search user profiles (async)
profiles = await client.search_profiles(
    request=SearchUserProfileRequest(
        user_id="your_user_id",
        query="john",
        threshold=0.7
    )
)
for profile in profiles.user_profiles:
    print(f"Profile {profile.profile_id}: {profile.profile_content}")

# Search interactions (async)
interactions = await client.search_interactions(
    request=SearchInteractionRequest(
        user_id="your_user_id",
        query="name is john"
    )
)
for interaction in interactions.interactions:
    print(f"Interaction {interaction.interaction_id}: {interaction.content}")

# Get all interactions (async)
all_interactions = await client.get_interactions(
    request={"user_id": "your_user_id", "top_k": 10}
)
for interaction in all_interactions.interactions:
    print(f"Interaction: {interaction.content}")

# Get all profiles (async)
all_profiles = await client.get_profiles(
    request={"user_id": "your_user_id", "top_k": 10}
)
for profile in all_profiles.user_profiles:
    print(f"Profile: {profile.profile_content}")

# Get profile change log (async)
change_log = await client.get_profile_change_log()
print(f"Profile changes: {len(change_log.profile_change_logs)} logs")

Features

Authentication

• Email/password login (async)
• API key authentication
• Custom endpoint configuration

User Interaction Management

• Publish user interactions (text, images, actions)
• Search interactions with semantic queries
• Get direct list of interactions
• Delete specific interactions
• Support for various interaction types (text, visual, click events)

User Profile Management

• Search user profiles with semantic queries
• Get direct list of user profiles
• Delete specific profiles or profiles matching search queries
• View profile change log history
• Filter by source and custom features

Configuration Management

• Get and set AgenticMem configuration
• Configure profile extractors
• Configure agent feedback systems
• Storage configuration (Supabase, S3, Local)

Agent Feedback System

• Get raw feedbacks from user interactions
• Get aggregated feedbacks
• Automatic feedback generation based on conversations
• Configurable feedback thresholds

API Response Types

All API methods return strongly-typed responses:

Authentication

• login() returns Token

Interaction Management

• publish_interaction() returns PublishUserInteractionResponse
• search_interactions() returns SearchInteractionResponse
• get_interactions() returns GetInteractionsResponse
• delete_interaction() returns DeleteUserInteractionResponse

Profile Management

• search_profiles() returns SearchUserProfileResponse
• get_profiles() returns GetUserProfilesResponse
• delete_profile() returns DeleteUserProfileResponse
• get_profile_change_log() returns ProfileChangeLogResponse

Configuration Management

• get_config() returns Config
• set_config() returns dict

Feedback Management

• get_raw_feedbacks() returns GetRawFeedbacksResponse
• get_feedbacks() returns GetFeedbacksResponse

Advanced Usage Examples

Visual Interactions with Images

import base64

# Function to encode the image
def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode("utf-8")

# Publish image interaction
base64_image = encode_image("./path/to/image.png")
await client.publish_interaction(
    request_id="your_request_id",
    user_id="your_user_id",
    interaction_requests=[
        InteractionRequest(
            content="I like this",
            image_encoding=base64_image,
        )
    ]
)

Configuration Management

from agenticmem import ProfileExtractorConfig, AgentFeedbackConfig, FeedbackAggregatorConfig

# Get current configuration
config = await client.get_config()
print(f"Current config: {config}")

# Update configuration
feedback_config = AgentFeedbackConfig(
    feedback_name="sales_feedback",
    feedback_definition_prompt="Extract actionable feedback for sales representatives",
    feedback_aggregator_config=FeedbackAggregatorConfig()
)

config.agent_feedback_configs = [feedback_config]
await client.set_config(config)

Feedback Management

# Get raw feedbacks
raw_feedbacks = await client.get_raw_feedbacks()
for feedback in raw_feedbacks.raw_feedbacks:
    print(f"Feedback: {feedback.feedback_content}")

# Get aggregated feedbacks
feedbacks = await client.get_feedbacks()
for feedback in feedbacks.feedbacks:
    print(f"Aggregated feedback: {feedback}")

Documentation

For detailed documentation, please visit the docs/ directory in this repository.

Important Notes

• All client methods are async and require await
• Image interactions support base64 encoded images
• The client supports flexible request types (dict or typed objects)
• Configuration changes affect the entire organization
• Feedback generation is automatic based on configured prompts

License

MIT License