zep
Zep | The Memory Foundation For Your AI Stack
Stars: 2446
Zep is a long-term memory service for AI Assistant apps. With Zep, you can provide AI assistants with the ability to recall past conversations, no matter how distant, while also reducing hallucinations, latency, and cost. Zep persists and recalls chat histories, and automatically generates summaries and other artifacts from these chat histories. It also embeds messages and summaries, enabling you to search Zep for relevant context from past conversations. Zep does all of this asyncronously, ensuring these operations don't impact your user's chat experience. Data is persisted to database, allowing you to scale out when growth demands. Zep also provides a simple, easy to use abstraction for document vector search called Document Collections. This is designed to complement Zep's core memory features, but is not designed to be a general purpose vector database. Zep allows you to be more intentional about constructing your prompt: 1. automatically adding a few recent messages, with the number customized for your app; 2. a summary of recent conversations prior to the messages above; 3. and/or contextually relevant summaries or messages surfaced from the entire chat session. 4. and/or relevant Business data from Zep Document Collections.
README:
Quick Start | Documentation | Zep Cloud Docs
Zep continually learns from user interactions, improving your AI agent's knowledge over time. With Zep, you can personalize user experiences and significantly improve agent accuracy.
Zep is powered by a temporal Knowledge Graph. As your user's conversation with an agent progresses, new facts are added to the graph. Zep maintains historical context, helping your agent reason with state change and offering data provenance insights.
Retrieving facts is simple and very fast. Both semantic and graph search are used to ensure facts are relevant to the current conversation. Fact retrieval does not require LLM inference; the slowest activity is embedding the search query.
Zep supports:
- Adding chat history messages.
- Ingestion of JSON and unstructured text.
- Session, user, and group-level graphs. Group graphs allow for capturing organizational knowledge.
Please see the Zep Quick Start Guide for important configuration information.
./zep pull
./zep up[!NOTE] Make sure to set the
secretvalue in thezep.yamlconfiguration file.Additionally, make sure that you expose an
OPENAI_API_KEYenvironment variable either in a local .env file or by runningexport OPENAI_API_KEY=your_openai_api_key
pip install zep-pythonor
npm i @getzep/zep-jsPersisting chat history memory is simple and fast.
result = await client.memory.add(session_id, messages=messages)Zep's high-level memory API offers an opinionated retrieval API, which uses BM25, semantic, and graph search to retrieve facts relevant to the current conversation. Results are reranked by distance from the user node, further improving relevance.
memory = client.memory.get(session_id="session_id")Lower-level APIs for search and CRUD are also available.
A Knowledge Graph is a network of interconnected facts, such as “Kendra loves Adidas shoes.” Each fact is a “triplet” represented by two entities, or nodes (”Kendra”, “Adidas shoes”), and their relationship, or edge (”loves”).
Knowledge Graphs allow us to model an agent's complex world and offer a superior retrieval approach than semantic search alone, which is commonly used in RAG. Most approaches to building Knowledge Graphs don't reason well with state changes. Facts inevitably change over time as users provide new information or business data changes.
Most graph-building tools don't reason well with state changes. Zep incorporates a temporal Knowledge Graph library, Graphiti, which we developed to address this challenge. What makes Graphiti unique is its ability to autonomously build a Knowledge Graph while handling changing relationships and maintaining historical context.
Graphiti also offers Zep the ability to ingest chat history, JSON business data, and unstructured text.
Zep is framework agnostic. It can be used with LangChain, LangGraph, Chainlit, Microsoft Autogen, and more.
Zep Community Edition is an open-source Zep distribution. It shares APIs with Zep Cloud and has comprehensive documentation available.
Zep Cloud is a managed service with Zep Community Edition at its core. In addition to Zep Community Edition's memory layer, Zep Cloud offers:
- Low Latency, Scalability, High Availability: Our cloud is designed to scale to the needs of customers with millions of DAUs and is SOC II Type 2 certified. Zep utilizes self-hosted LLMs and embedding models, offering customers very low-latency memory retrieval and graph-building.
- Dialog Classification: Instantly and accurately classify chat dialog. Understand user intent and emotion, segment users, and more. Route chains based on semantic context, and trigger events.
- Structured Data Extraction: Quickly extract business data from chat conversations using a schema you define. Understand what your assistant should ask for next to complete the task.
With increased LLM context lengths, including the entire chat history, RAG results, and other instructions in a prompt may be tempting. Unfortunately, this has resulted in poor temporal reasoning, poor recall, hallucinations, and slow and expensive inference.
As discussed above, providing just the chat history to an LLM can often result in poor temporal reasoning.
Users, Sessions, and Chat Messages are first-class abstractions in Zep. This allows simple and flexible management of chat memory, including the execution of Right To Be Forgetten requests and other privacy compliance-related tasks with single-API call.
Yes - Zep offers Python & TypeScript/JS SDKs for easy integration with your Assistant app. We also have examples of using Zep with popular frameworks - see below.
Yes - the Zep team and community contributors have built integrations with Zep, making it simple to, for example, drop Zep's memory components into a LangChain app. Please see the Zep Documentation and your favorite framework's documentation.
Zep Community Edition relies on an external LLM API service to function. Any OpenAI-compatible LLM API is supported. Providers such as Anthropic can be used via a proxy such as LiteLLM. You will also need to configure LiteLLM with an embedding service.
import uuid
from zep_python.client import AsyncZep
from zep_python.types import Message
client = AsyncZep(
api_key=API_KEY,
base_url=BASE_URL,
)
user_id = uuid.uuid4().hex # A new user identifier
new_user = await client.user.add(
user_id=user_id,
email="[email protected]",
first_name="Jane",
last_name="Smith",
metadata={"foo": "bar"},
)
# create a chat session
session_id = uuid.uuid4().hex # A new session identifier
session = await client.memory.add_session(
session_id=session_id,
user_id=user_id,
metadata={"foo" : "bar"}
)
# Add a memory to the session
await client.memory.add_memory(
session_id=session_id,
messages=[
Message(
role_type = "user", # One of ("system", "assistant", "user", "function", "tool")
role = "Researcher", # Optional, a use case specific string representing the role of the user
content = "Who was Octavia Butler?", # The message content
)
],
)
# Get session memory
memory = await client.memory.get(session_id=session_id)
messages = memory.messages # List of messages in the session (quantity determined by optional lastn parameter in memory.get)
relevant_facts = memory.relevant_facts # List of facts relevant to the recent messages in the session
# Search user facts across all sessions
search_response = await client.memory.search_sessions(
user_id=user_id,
search_scope="facts",
text="What science fiction books did I recently read?",
)
facts = [r.fact for r in search_response.results]import { v4 as uuidv4 } from 'uuid';
import { ZepClient } from '@getzep/zep-js';
import type { CreateUserRequest, CreateSessionRequest, SessionSearchQuery } from '@getzep/zep-js/api';
const client = new ZepClient({
apiKey: API_KEY,
baseUrl: BASE_URL,
});
// A new user identifier
const userId = uuidv4();
const userRequest: CreateUserRequest = {
userId: userId,
email: "[email protected]",
firstName: "Jane",
lastName: "Smith",
metadata: { foo: "bar" },
};
const newUser = await client.user.add(userRequest);
// Create a chat session
const sessionId = uuidv4();
const sessionRequest: CreateSessionRequest = {
sessionId: sessionId,
userId: userId,
metadata: { foo: "bar" },
};
// A new session identifier
const session = await client.memory.addSession(sessionRequest);
// Add a memory to the session
await client.memory.add(sessionId, {
messages: [
{
role: "Researcher",
roleType: "user",
content: "Who was Octavia Butler?",
},
],
});
// Get session memory
const memory = await client.memory.get(sessionId);
const messages = memory.messages; // List of messages in the session (quantity determined by optional lastN parameter in memory.get)
const relevantFacts = memory.relevantFacts; // List of facts relevant to the recent messages in the session
// Search user facts across all sessions
const searchQuery: SessionSearchQuery = {
userId: userId,
searchScope: "facts",
text: "What science fiction books did I recently read?",
};
const searchResponse = await client.memory.searchSessions(searchQuery);
const facts = searchResponse.results?.map(result => result.fact);Zep Open Source is an older version of Zep that did not use a Knowledge Graph to persist and recall memory.
Some additional changes:
- The Zep OSS web UI has been deprecated in favor of significantly expanded SDK support.
- Zep CE supports many LLM services and local servers that offer OpenAI-compatible APIs. Other services may be used with an LLM proxy.
- Zep CE no longer ships with a local embedding service and named entity extractor.
Significant changes have been made to Zep, and unfortunately, we have not been able to devise a migration path from Zep OSS to Zep CE.
Zep OSS will remain available in our container repo, but we will not see future enhancements or bug fixes. The code is available in the legacy branch in this repo.
We welcome contributions. For more, see the CONTRIBUTING file in this repo.
For Tasks:
Click tags to check more tools for each tasksFor Jobs:
Alternative AI tools for zep
Similar Open Source Tools
zep
Zep is a long-term memory service for AI Assistant apps. With Zep, you can provide AI assistants with the ability to recall past conversations, no matter how distant, while also reducing hallucinations, latency, and cost. Zep persists and recalls chat histories, and automatically generates summaries and other artifacts from these chat histories. It also embeds messages and summaries, enabling you to search Zep for relevant context from past conversations. Zep does all of this asyncronously, ensuring these operations don't impact your user's chat experience. Data is persisted to database, allowing you to scale out when growth demands. Zep also provides a simple, easy to use abstraction for document vector search called Document Collections. This is designed to complement Zep's core memory features, but is not designed to be a general purpose vector database. Zep allows you to be more intentional about constructing your prompt: 1. automatically adding a few recent messages, with the number customized for your app; 2. a summary of recent conversations prior to the messages above; 3. and/or contextually relevant summaries or messages surfaced from the entire chat session. 4. and/or relevant Business data from Zep Document Collections.
langchain
LangChain is a framework for developing Elixir applications powered by language models. It enables applications to connect language models to other data sources and interact with the environment. The library provides components for working with language models and off-the-shelf chains for specific tasks. It aims to assist in building applications that combine large language models with other sources of computation or knowledge. LangChain is written in Elixir and is not aimed for parity with the JavaScript and Python versions due to differences in programming paradigms and design choices. The library is designed to make it easy to integrate language models into applications and expose features, data, and functionality to the models.
embedchain
Embedchain is an Open Source Framework for personalizing LLM responses. It simplifies the creation and deployment of personalized AI applications by efficiently managing unstructured data, generating relevant embeddings, and storing them in a vector database. With diverse APIs, users can extract contextual information, find precise answers, and engage in interactive chat conversations tailored to their data. The framework follows the design principle of being 'Conventional but Configurable' to cater to both software engineers and machine learning engineers.
sdk
Vikit.ai SDK is a software development kit that enables easy development of video generators using generative AI and other AI models. It serves as a langchain to orchestrate AI models and video editing tools. The SDK allows users to create videos from text prompts with background music and voice-over narration. It also supports generating composite videos from multiple text prompts. The tool requires Python 3.8+, specific dependencies, and tools like FFMPEG and ImageMagick for certain functionalities. Users can contribute to the project by following the contribution guidelines and standards provided.
kafka-ml
Kafka-ML is a framework designed to manage the pipeline of Tensorflow/Keras and PyTorch machine learning models on Kubernetes. It enables the design, training, and inference of ML models with datasets fed through Apache Kafka, connecting them directly to data streams like those from IoT devices. The Web UI allows easy definition of ML models without external libraries, catering to both experts and non-experts in ML/AI.
llms
The 'llms' repository is a comprehensive guide on Large Language Models (LLMs), covering topics such as language modeling, applications of LLMs, statistical language modeling, neural language models, conditional language models, evaluation methods, transformer-based language models, practical LLMs like GPT and BERT, prompt engineering, fine-tuning LLMs, retrieval augmented generation, AI agents, and LLMs for computer vision. The repository provides detailed explanations, examples, and tools for working with LLMs.
Woodpecker
Woodpecker is a tool designed to correct hallucinations in Multimodal Large Language Models (MLLMs) by introducing a training-free method that picks out and corrects inconsistencies between generated text and image content. It consists of five stages: key concept extraction, question formulation, visual knowledge validation, visual claim generation, and hallucination correction. Woodpecker can be easily integrated with different MLLMs and provides interpretable results by accessing intermediate outputs of the stages. The tool has shown significant improvements in accuracy over baseline models like MiniGPT-4 and mPLUG-Owl.
studio-b3
Studio B3 (B-3 Bomber) is a sophisticated editor designed for content creation, catering to various formats such as blogs, articles, user stories, and more. It provides an immersive content generation experience with local AI capabilities for intelligent search and recommendation functions. Users can define custom actions and variables for flexible content generation. The editor includes interactive tools like Bubble Menu, Slash Command, and Quick Insert for enhanced user experience in editing, searching, and navigation. The design principles focus on intelligent embedding of AI, local optimization for efficient writing experience, and context flexibility for better control over AI-generated content.
LangChain
LangChain is a C# implementation of the LangChain library, which provides a composable way to build applications with LLMs (Large Language Models). It offers a variety of features, including: - A unified interface for interacting with different LLMs, such as OpenAI's GPT-3 and Microsoft's Azure OpenAI Service - A set of pre-built chains that can be used to perform common tasks, such as question answering, summarization, and translation - A flexible API that allows developers to create their own custom chains - A growing community of developers and users who are contributing to the project LangChain is still under development, but it is already being used to build a variety of applications, including chatbots, search engines, and writing assistants. As the project continues to mature, it is expected to become an increasingly valuable tool for developers who want to build applications with LLMs.
datasets
Datasets is a repository that provides a collection of various datasets for machine learning and data analysis projects. It includes datasets in different formats such as CSV, JSON, and Excel, covering a wide range of topics including finance, healthcare, marketing, and more. The repository aims to help data scientists, researchers, and students access high-quality datasets for training models, conducting experiments, and exploring data analysis techniques.
mastra
Mastra is an opinionated Typescript framework designed to help users quickly build AI applications and features. It provides primitives such as workflows, agents, RAG, integrations, syncs, and evals. Users can run Mastra locally or deploy it to a serverless cloud. The framework supports various LLM providers, offers tools for building language models, workflows, and accessing knowledge bases. It includes features like durable graph-based state machines, retrieval-augmented generation, integrations, syncs, and automated tests for evaluating LLM outputs.
numerapi
Numerapi is a Python client to the Numerai API that allows users to automatically download and upload data for the Numerai machine learning competition. It provides functionalities for downloading training data, uploading predictions, and accessing user, submission, and competitions information for both the main competition and Numerai Signals competition. Users can interact with the API using Python modules or command line interface. Tokens are required for certain actions like uploading predictions or staking, which can be obtained from Numer.ai account settings. The tool also supports features like checking new rounds, getting leaderboards, and managing stakes.
langgraphjs
LangGraph.js is a library for building stateful, multi-actor applications with LLMs, offering benefits such as cycles, controllability, and persistence. It allows defining flows involving cycles, providing fine-grained control over application flow and state. Inspired by Pregel and Apache Beam, it includes features like loops, persistence, human-in-the-loop workflows, and streaming support. LangGraph integrates seamlessly with LangChain.js and LangSmith but can be used independently.
Trinity
Trinity is an Explainable AI (XAI) Analysis and Visualization tool designed for Deep Learning systems or other models performing complex classification or decoding. It provides performance analysis through interactive 3D projections that are hyper-dimensional aware, allowing users to explore hyperspace, hypersurface, projections, and manifolds. Trinity primarily works with JSON data formats and supports the visualization of FeatureVector objects. Users can analyze and visualize data points, correlate inputs with classification results, and create custom color maps for better data interpretation. Trinity has been successfully applied to various use cases including Deep Learning Object detection models, COVID gene/tissue classification, Brain Computer Interface decoders, and Large Language Model (ChatGPT) Embeddings Analysis.
hi-ml
The Microsoft Health Intelligence Machine Learning Toolbox is a repository that provides low-level and high-level building blocks for Machine Learning / AI researchers and practitioners. It simplifies and streamlines work on deep learning models for healthcare and life sciences by offering tested components such as data loaders, pre-processing tools, deep learning models, and cloud integration utilities. The repository includes two Python packages, 'hi-ml-azure' for helper functions in AzureML, 'hi-ml' for ML components, and 'hi-ml-cpath' for models and workflows related to histopathology images.
keras-hub
KerasHub is a pretrained modeling library that provides Keras 3 implementations of popular model architectures with pretrained checkpoints. It supports text, image, and audio data for generation, classification, and other tasks. Models are compatible with JAX, TensorFlow, and PyTorch, and can be fine-tuned on GPUs and TPUs. KerasHub components are provided as Layer and Model implementations, extending the core Keras API.
For similar tasks
zep
Zep is a long-term memory service for AI Assistant apps. With Zep, you can provide AI assistants with the ability to recall past conversations, no matter how distant, while also reducing hallucinations, latency, and cost. Zep persists and recalls chat histories, and automatically generates summaries and other artifacts from these chat histories. It also embeds messages and summaries, enabling you to search Zep for relevant context from past conversations. Zep does all of this asyncronously, ensuring these operations don't impact your user's chat experience. Data is persisted to database, allowing you to scale out when growth demands. Zep also provides a simple, easy to use abstraction for document vector search called Document Collections. This is designed to complement Zep's core memory features, but is not designed to be a general purpose vector database. Zep allows you to be more intentional about constructing your prompt: 1. automatically adding a few recent messages, with the number customized for your app; 2. a summary of recent conversations prior to the messages above; 3. and/or contextually relevant summaries or messages surfaced from the entire chat session. 4. and/or relevant Business data from Zep Document Collections.
classifai
Supercharge WordPress Content Workflows and Engagement with Artificial Intelligence. Tap into leading cloud-based services like OpenAI, Microsoft Azure AI, Google Gemini and IBM Watson to augment your WordPress-powered websites. Publish content faster while improving SEO performance and increasing audience engagement. ClassifAI integrates Artificial Intelligence and Machine Learning technologies to lighten your workload and eliminate tedious tasks, giving you more time to create original content that matters.
ontogpt
OntoGPT is a Python package for extracting structured information from text using large language models, instruction prompts, and ontology-based grounding. It provides a command line interface and a minimal web app for easy usage. The tool has been evaluated on test data and is used in related projects like TALISMAN for gene set analysis. OntoGPT enables users to extract information from text by specifying relevant terms and provides the extracted objects as output.
mslearn-ai-language
This repository contains lab files for Azure AI Language modules. It provides hands-on exercises and resources for learning about various AI language technologies on the Azure platform. The labs cover topics such as natural language processing, text analytics, language understanding, and more. By following the exercises in this repository, users can gain practical experience in implementing AI language solutions using Azure services.
summary-of-a-haystack
This repository contains data and code for the experiments in the SummHay paper. It includes publicly released Haystacks in conversational and news domains, along with scripts for running the pipeline, visualizing results, and benchmarking automatic evaluation. The data structure includes topics, subtopics, insights, queries, retrievers, summaries, evaluation summaries, and documents. The pipeline involves scripts for retriever scores, summaries, and evaluation scores using GPT-4o. Visualization scripts are provided for compiling and visualizing results. The repository also includes annotated samples for benchmarking and citation information for the SummHay paper.
llm-book
The 'llm-book' repository is dedicated to the introduction of large-scale language models, focusing on natural language processing tasks. The code is designed to run on Google Colaboratory and utilizes datasets and models available on the Hugging Face Hub. Note that as of July 28, 2023, there are issues with the MARC-ja dataset links, but an alternative notebook using the WRIME Japanese sentiment analysis dataset has been added. The repository covers various chapters on topics such as Transformers, fine-tuning language models, entity recognition, summarization, document embedding, question answering, and more.
Controllable-RAG-Agent
This repository contains a sophisticated deterministic graph-based solution for answering complex questions using a controllable autonomous agent. The solution is designed to ensure that answers are solely based on the provided data, avoiding hallucinations. It involves various steps such as PDF loading, text preprocessing, summarization, database creation, encoding, and utilizing large language models. The algorithm follows a detailed workflow involving planning, retrieval, answering, replanning, content distillation, and performance evaluation. Heuristics and techniques implemented focus on content encoding, anonymizing questions, task breakdown, content distillation, chain of thought answering, verification, and model performance evaluation.
summarize
The 'summarize' tool is designed to transcribe and summarize videos from various sources using AI models. It helps users efficiently summarize lengthy videos, take notes, and extract key insights by providing timestamps, original transcripts, and support for auto-generated captions. Users can utilize different AI models via Groq, OpenAI, or custom local models to generate grammatically correct video transcripts and extract wisdom from video content. The tool simplifies the process of summarizing video content, making it easier to remember and reference important information.
For similar jobs
zep
Zep is a long-term memory service for AI Assistant apps. With Zep, you can provide AI assistants with the ability to recall past conversations, no matter how distant, while also reducing hallucinations, latency, and cost. Zep persists and recalls chat histories, and automatically generates summaries and other artifacts from these chat histories. It also embeds messages and summaries, enabling you to search Zep for relevant context from past conversations. Zep does all of this asyncronously, ensuring these operations don't impact your user's chat experience. Data is persisted to database, allowing you to scale out when growth demands. Zep also provides a simple, easy to use abstraction for document vector search called Document Collections. This is designed to complement Zep's core memory features, but is not designed to be a general purpose vector database. Zep allows you to be more intentional about constructing your prompt: 1. automatically adding a few recent messages, with the number customized for your app; 2. a summary of recent conversations prior to the messages above; 3. and/or contextually relevant summaries or messages surfaced from the entire chat session. 4. and/or relevant Business data from Zep Document Collections.
doc2plan
doc2plan is a browser-based application that helps users create personalized learning plans by extracting content from documents. It features a Creator for manual or AI-assisted plan construction and a Viewer for interactive plan navigation. Users can extract chapters, key topics, generate quizzes, and track progress. The application includes AI-driven content extraction, quiz generation, progress tracking, plan import/export, assistant management, customizable settings, viewer chat with text-to-speech and speech-to-text support, and integration with various Retrieval-Augmented Generation (RAG) models. It aims to simplify the creation of comprehensive learning modules tailored to individual needs.
whatsapp-chatgpt
This repository contains a WhatsApp bot that utilizes OpenAI's GPT and DALL-E 2 to respond to user inputs. Users can interact with the bot through voice messages, which are transcribed and responded to. The bot requires Node.js, npm, an OpenAI API key, and a WhatsApp account. It uses Puppeteer to run a real instance of Whatsapp Web to avoid being blocked. However, there is a risk of being blocked by WhatsApp as it does not allow bots or unofficial clients on its platform. The bot is not free to use, and users will be charged by OpenAI for each request made.
OmniSteward
OmniSteward is an AI-powered steward system based on large language models that can interact with users through voice or text to help control smart home devices and computer programs. It supports multi-turn dialogue, tool calling for complex tasks, multiple LLM models, voice recognition, smart home control, computer program management, online information retrieval, command line operations, and file management. The system is highly extensible, allowing users to customize and share their own tools.
chatgpt-wechat
ChatGPT-WeChat is a personal assistant application that can be safely used on WeChat through enterprise WeChat without the risk of being banned. The project is open source and free, with no paid sections or external traffic operations except for advertising on the author's public account '积木成楼'. It supports various features such as secure usage on WeChat, multi-channel customer service message integration, proxy support, session management, rapid message response, voice and image messaging, drawing capabilities, private data storage, plugin support, and more. Users can also develop their own capabilities following the rules provided. The project is currently in development with stable versions available for use.
mcp-agent
mcp-agent is a simple, composable framework designed to build agents using the Model Context Protocol. It handles the lifecycle of MCP server connections and implements patterns for building production-ready AI agents in a composable way. The framework also includes OpenAI's Swarm pattern for multi-agent orchestration in a model-agnostic manner, making it the simplest way to build robust agent applications. It is purpose-built for the shared protocol MCP, lightweight, and closer to an agent pattern library than a framework. mcp-agent allows developers to focus on the core business logic of their AI applications by handling mechanics such as server connections, working with LLMs, and supporting external signals like human input.
Gmail-MCP-Server
Gmail AutoAuth MCP Server is a Model Context Protocol (MCP) server designed for Gmail integration in Claude Desktop. It supports auto authentication and enables AI assistants to manage Gmail through natural language interactions. The server provides comprehensive features for sending emails, reading messages, managing labels, searching emails, and batch operations. It offers full support for international characters, email attachments, and Gmail API integration. Users can install and authenticate the server via Smithery or manually with Google Cloud Project credentials. The server supports both Desktop and Web application credentials, with global credential storage for convenience. It also includes Docker support and instructions for cloud server authentication.
Operit
Operit AI is a fully functional AI assistant application for mobile devices, running independently on Android devices with powerful tool invocation capabilities. It offers over 40 built-in tools for file system operations, HTTP requests, system operations, UI automation, and media processing. The app combines these tools with rich plugins to enable a wide range of tasks, from simple to complex, providing a comprehensive experience of a smartphone AI assistant.