{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"phidata","slug":"phidata","name":"Phidata","type":"framework","url":"https://github.com/phidatahq/phidata","page_url":"https://unfragile.ai/phidata","categories":["ai-agents","rag-knowledge"],"tags":[],"pricing":{"model":"free","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"phidata__cap_0","uri":"capability://tool.use.integration.multi.provider.llm.abstraction.with.unified.interface","name":"multi-provider llm abstraction with unified interface","description":"Provides a unified Python API that abstracts across OpenAI, Anthropic, Google, Ollama, and other LLM providers through a common Agent class. Internally routes requests to provider-specific SDK clients while normalizing request/response formats, enabling seamless provider switching without code changes. Handles model-specific parameter mapping (e.g., temperature, max_tokens) and response parsing across different API schemas.","intents":["I want to build an agent that works with multiple LLM providers without rewriting code for each one","I need to switch between OpenAI and Anthropic models at runtime based on cost or availability","I want to use local models (Ollama) in development and cloud models (OpenAI) in production with the same agent code"],"best_for":["teams building multi-model applications to avoid vendor lock-in","developers prototyping with different LLMs to find optimal cost/performance tradeoffs","enterprises requiring on-premise model support alongside cloud APIs"],"limitations":["Provider-specific features (e.g., vision capabilities, function calling schemas) may not be fully normalized across all providers","Response latency varies significantly by provider; no built-in load balancing or failover between providers","Some advanced parameters (e.g., Anthropic's thinking budget) may not map cleanly to the unified interface"],"requires":["Python 3.9+","API keys for at least one provider (OpenAI, Anthropic, Google, etc.) OR local Ollama instance","phidata package installed via pip"],"input_types":["text prompts","system instructions","structured parameters (temperature, max_tokens, top_p)"],"output_types":["text responses","structured JSON (when using function calling)","streaming token chunks"],"categories":["tool-use-integration","abstraction-layer"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_1","uri":"capability://tool.use.integration.function.calling.with.schema.based.tool.registration","name":"function calling with schema-based tool registration","description":"Enables agents to invoke external functions through a schema-based tool registry that automatically generates OpenAI/Anthropic-compatible function schemas from Python function signatures and docstrings. The framework handles schema generation, function invocation, and response parsing, supporting both synchronous and asynchronous tool execution. Tools are registered declaratively and the agent automatically includes them in function_calling requests to the LLM.","intents":["I want my agent to call Python functions, APIs, or databases based on LLM reasoning without manual prompt engineering","I need to expose a set of tools to an agent and have it decide when and how to use them","I want to build a multi-step workflow where the agent chains tool calls together to solve complex problems"],"best_for":["developers building autonomous agents that need to interact with external systems","teams creating domain-specific agents with custom business logic tools","builders prototyping agentic workflows without writing complex prompt templates"],"limitations":["Schema generation from Python functions may fail for complex types (nested generics, custom classes); requires explicit Pydantic models for reliability","No built-in retry logic for failed tool calls; requires manual implementation of error handling and recovery","Async tool execution requires careful management of event loops; blocking tools will stall the entire agent"],"requires":["Python 3.9+","LLM provider that supports function calling (OpenAI, Anthropic, Google, etc.)","Pydantic for type hints (optional but recommended for complex schemas)"],"input_types":["Python functions with type hints","Pydantic models for tool parameters","docstrings describing tool behavior"],"output_types":["function execution results","structured tool responses","error messages from failed tool calls"],"categories":["tool-use-integration","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_10","uri":"capability://planning.reasoning.custom.agent.reasoning.with.chain.of.thought.prompting","name":"custom agent reasoning with chain-of-thought prompting","description":"Enables agents to use chain-of-thought reasoning patterns where the LLM explicitly breaks down problems into steps before generating final answers. The framework automatically constructs prompts that encourage step-by-step reasoning, captures intermediate reasoning steps, and uses them to improve final outputs. Supports both explicit chain-of-thought (shown to users) and implicit reasoning (internal only).","intents":["I want my agent to show its reasoning process to users for transparency","I need agents to break down complex problems into steps for better accuracy","I want to improve agent reasoning quality by encouraging explicit problem decomposition"],"best_for":["developers building agents that need to explain their reasoning","teams creating educational or transparent AI systems","applications where reasoning quality is critical (math, logic, analysis)"],"limitations":["Chain-of-thought adds latency; LLM must generate reasoning steps before final answer","Reasoning quality depends on LLM capability; weaker models may produce poor reasoning","No automatic verification of reasoning correctness; agents may show confident but incorrect reasoning","Reasoning steps may be verbose or confusing to users; requires careful prompt engineering"],"requires":["Python 3.9+","LLM capable of multi-step reasoning (GPT-4, Claude, etc.)"],"input_types":["complex problems","reasoning prompts"],"output_types":["reasoning steps","final answers","confidence scores"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_11","uri":"capability://text.generation.language.custom.system.prompts.and.agent.personality.configuration","name":"custom system prompts and agent personality configuration","description":"Allows customization of agent behavior through system prompts and personality configuration. Developers can define custom instructions, constraints, tone, and behavioral guidelines that shape how agents respond. System prompts are automatically prepended to all LLM calls, ensuring consistent agent behavior across interactions. Supports prompt templates with variable substitution for dynamic configuration.","intents":["I want to customize how my agent behaves and responds to users","I need to enforce specific constraints or guidelines on agent behavior","I want to create agents with distinct personalities or communication styles"],"best_for":["developers building domain-specific agents with custom behavior","teams creating branded AI assistants with consistent personalities","applications requiring agents to follow specific guidelines or constraints"],"limitations":["System prompts are not guaranteed to be followed; LLMs may ignore or violate instructions","Prompt engineering is an art; small changes to system prompts can significantly affect behavior","No automatic validation that agents follow system prompt constraints; requires manual testing","Complex system prompts may confuse the LLM or reduce response quality"],"requires":["Python 3.9+","Understanding of prompt engineering best practices"],"input_types":["system prompt text","personality configuration","constraint definitions"],"output_types":["customized agent behavior","responses aligned with system prompt"],"categories":["text-generation-language","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_12","uri":"capability://data.processing.analysis.document.processing.and.chunking.for.knowledge.ingestion","name":"document processing and chunking for knowledge ingestion","description":"Provides utilities for processing various document formats (PDF, markdown, plain text, web pages) and chunking them into manageable pieces for embedding and retrieval. Handles document parsing, text extraction, metadata preservation, and intelligent chunking strategies (semantic, fixed-size, sliding window). Chunks are automatically embedded and stored in knowledge bases for RAG.","intents":["I want to ingest a large document corpus into a knowledge base for RAG","I need to extract text from PDFs and other formats for agent knowledge","I want to chunk documents intelligently to preserve semantic meaning"],"best_for":["teams building knowledge bases from document collections","developers creating document-based Q&A systems","applications requiring intelligent document processing"],"limitations":["PDF parsing is fragile; complex layouts, images, or scanned documents may fail to extract correctly","Chunking strategy significantly impacts retrieval quality; no automatic optimization of chunk size","Metadata extraction is limited; requires manual configuration for domain-specific metadata","Large document collections require significant processing time and storage"],"requires":["Python 3.9+","Document files in supported formats (PDF, markdown, text, HTML)","Optional: PDF parsing libraries (PyPDF2, pdfplumber)"],"input_types":["document files","chunking configuration (chunk size, overlap, strategy)"],"output_types":["text chunks","metadata","embeddings"],"categories":["data-processing-analysis","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_13","uri":"capability://image.visual.vision.capabilities.for.image.analysis.and.understanding","name":"vision capabilities for image analysis and understanding","description":"Phidata integrates vision models (OpenAI Vision, Claude Vision, etc.) for analyzing images and providing detailed descriptions, object detection, text extraction (OCR), and visual reasoning. The framework handles image encoding, provider-specific vision API calls, and response parsing for vision-enabled agents.","intents":["I want agents to analyze images and answer questions about visual content","I need to extract text from images (OCR) using LLM vision capabilities","I want to build agents that can process both text and image inputs"],"best_for":["Document processing applications requiring OCR","Visual question-answering systems","Multi-modal agents handling text and images"],"limitations":["Vision model accuracy varies; complex images may be misinterpreted","Image encoding and transmission adds latency (500ms-2s per image)","Not all providers support vision equally; feature parity is inconsistent","Large images may be resized or compressed, losing detail"],"requires":["Python 3.8+","Vision-enabled LLM provider (OpenAI, Anthropic, Google)","Image files (JPEG, PNG, GIF, WebP)"],"input_types":["image files (JPEG, PNG, GIF, WebP)","image URLs","image bytes"],"output_types":["image descriptions (text)","extracted text (OCR)","visual analysis (structured data)","answers to visual questions (text)"],"categories":["image-visual","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_2","uri":"capability://memory.knowledge.agent.memory.with.session.persistence","name":"agent memory with session persistence","description":"Provides a pluggable memory system that stores conversation history, tool call results, and agent state across sessions. Supports multiple backends (in-memory, SQLite, PostgreSQL) and automatically manages message history, context windows, and memory summarization. Memory is attached to agents and automatically updated after each interaction, enabling stateful multi-turn conversations and long-running agent instances.","intents":["I want my agent to remember previous conversations and context across multiple interactions","I need to persist agent state so it survives application restarts","I want to implement memory summarization or pruning to manage context window limits"],"best_for":["developers building conversational agents that need multi-turn context","teams deploying long-running agents that must survive restarts","applications requiring audit trails of agent decisions and tool calls"],"limitations":["In-memory storage is lost on process termination; requires explicit persistence layer for production","No built-in memory summarization or compression; full history grows linearly with conversation length","SQLite backend is single-process; PostgreSQL required for multi-instance deployments","Memory retrieval is sequential; no semantic search or relevance ranking across stored messages"],"requires":["Python 3.9+","SQLite (built-in) or PostgreSQL for persistent storage","phidata package with memory module"],"input_types":["conversation messages (user, assistant, tool)","tool call results","custom metadata"],"output_types":["message history","conversation summaries","memory statistics"],"categories":["memory-knowledge","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_3","uri":"capability://memory.knowledge.rag.retrieval.augmented.generation.with.knowledge.base.integration","name":"rag (retrieval-augmented generation) with knowledge base integration","description":"Integrates vector-based retrieval with agents through a Knowledge class that chunks documents, generates embeddings, and stores them in vector databases (Pinecone, Weaviate, Chroma, etc.). Agents can retrieve relevant documents before generating responses, augmenting their knowledge with external sources. The framework handles embedding generation, similarity search, and result ranking automatically.","intents":["I want my agent to answer questions based on a large document corpus without fine-tuning","I need to ground agent responses in specific documents to reduce hallucinations","I want to build a Q&A agent over proprietary data (PDFs, websites, databases)"],"best_for":["teams building domain-specific Q&A systems over proprietary documents","developers creating customer support agents grounded in knowledge bases","enterprises needing to augment LLMs with internal data without fine-tuning"],"limitations":["Embedding quality depends on the embedding model; no automatic tuning of chunk size or overlap for optimal retrieval","Vector database integration requires external service setup (Pinecone, Weaviate) or local instance (Chroma); adds operational complexity","Retrieval is purely semantic; no keyword search or hybrid retrieval fallback for rare terms","No built-in evaluation of retrieval quality; requires manual testing to verify relevant documents are returned"],"requires":["Python 3.9+","Embedding model (OpenAI, Hugging Face, or local)","Vector database (Pinecone, Weaviate, Chroma, etc.) or local SQLite with embeddings","Documents in supported formats (PDF, text, markdown, etc.)"],"input_types":["documents (PDF, text, markdown, web pages)","user queries","embedding model selection"],"output_types":["retrieved document chunks","similarity scores","augmented prompts with context"],"categories":["memory-knowledge","search-retrieval"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_4","uri":"capability://data.processing.analysis.structured.output.generation.with.pydantic.models","name":"structured output generation with pydantic models","description":"Enables agents to generate structured, validated outputs by specifying Pydantic models as response schemas. The framework automatically constructs prompts that guide the LLM to produce JSON matching the schema, then parses and validates responses. Supports nested models, optional fields, and custom validators, ensuring type-safe agent outputs.","intents":["I want my agent to return structured data (JSON) instead of free-form text","I need to validate agent outputs against a schema before using them in downstream systems","I want to build agents that generate structured reports, extractions, or classifications"],"best_for":["developers building agents that feed into structured APIs or databases","teams creating data extraction or classification agents","applications requiring type-safe agent outputs for downstream processing"],"limitations":["LLM may fail to generate valid JSON matching the schema; requires fallback parsing or retry logic","Complex nested schemas may confuse the LLM; schema design significantly impacts output quality","No built-in schema optimization; developers must manually simplify schemas to improve LLM compliance","Validation errors are not automatically recovered; failed outputs require manual intervention or retry"],"requires":["Python 3.9+","Pydantic v1 or v2","LLM provider supporting structured outputs (OpenAI, Anthropic, Google)"],"input_types":["Pydantic model definitions","user prompts","context data"],"output_types":["validated Python objects","JSON strings","validation errors"],"categories":["data-processing-analysis","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_5","uri":"capability://planning.reasoning.multi.agent.orchestration.and.team.workflows","name":"multi-agent orchestration and team workflows","description":"Supports building teams of specialized agents that collaborate to solve complex problems. Agents can delegate tasks to other agents, share memory and knowledge bases, and coordinate through a supervisor or hierarchical structure. The framework provides patterns for agent-to-agent communication, result aggregation, and workflow coordination without requiring manual message passing.","intents":["I want to build a system where multiple specialized agents work together on a complex task","I need agents to delegate subtasks to other agents and aggregate results","I want to create a hierarchical team structure where a supervisor agent coordinates worker agents"],"best_for":["teams building complex multi-step workflows that benefit from agent specialization","developers creating supervisor-worker agent architectures","applications requiring coordination between agents with different capabilities"],"limitations":["No built-in load balancing or resource management across agents; all agents run in the same process","Agent-to-agent communication is synchronous; no async delegation or parallel execution of independent tasks","No built-in conflict resolution when agents disagree; requires manual implementation of consensus logic","Debugging multi-agent workflows is complex; no built-in tracing or visualization of agent interactions"],"requires":["Python 3.9+","Multiple agent instances with distinct roles/capabilities","Shared memory or knowledge base for agent coordination"],"input_types":["user queries","task descriptions","agent configurations"],"output_types":["aggregated results from multiple agents","delegation logs","final responses"],"categories":["planning-reasoning","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_6","uri":"capability://text.generation.language.streaming.response.generation.with.token.level.control","name":"streaming response generation with token-level control","description":"Enables agents to stream responses token-by-token to clients, providing real-time feedback and reducing perceived latency. The framework handles streaming protocol negotiation with different LLM providers, token buffering, and graceful error handling during streams. Supports both text streaming and structured output streaming with partial validation.","intents":["I want to stream agent responses to users in real-time instead of waiting for complete generation","I need to display tokens as they arrive to improve perceived performance","I want to implement cancellation or early stopping during agent response generation"],"best_for":["developers building interactive chat applications with agents","teams creating real-time AI interfaces where latency perception matters","applications requiring token-level control for cost monitoring or rate limiting"],"limitations":["Streaming breaks structured output validation; partial JSON cannot be validated until complete","Error handling during streams is complex; connection drops may leave clients with incomplete responses","Token counting is approximate; actual token usage may differ from streamed token count","Streaming adds complexity to agent logic; not all agent patterns (e.g., multi-agent coordination) work well with streaming"],"requires":["Python 3.9+","LLM provider supporting streaming (OpenAI, Anthropic, Google, Ollama)","Client capable of consuming streaming responses (WebSocket, Server-Sent Events, etc.)"],"input_types":["prompts","streaming configuration (buffer size, timeout)"],"output_types":["token streams","partial responses","stream metadata (tokens, latency)"],"categories":["text-generation-language","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_7","uri":"capability://automation.workflow.agent.monitoring.and.logging.with.execution.traces","name":"agent monitoring and logging with execution traces","description":"Provides built-in logging and monitoring of agent execution, capturing all LLM calls, tool invocations, memory updates, and decision points. Generates detailed execution traces that show the agent's reasoning path, including prompts sent to the LLM, responses received, and tool results. Traces can be exported for debugging, auditing, or performance analysis.","intents":["I want to debug why an agent made a particular decision or took a specific action","I need to audit all agent decisions and tool calls for compliance or troubleshooting","I want to analyze agent performance and identify bottlenecks in the execution pipeline"],"best_for":["teams deploying agents in production and needing observability","developers debugging complex agent behaviors","enterprises requiring audit trails of agent decisions for compliance"],"limitations":["Detailed logging adds overhead; high-volume agent execution may impact performance","Traces can be very large for long conversations; storage and retrieval of traces requires external systems","No built-in analysis or visualization of traces; requires custom tooling to extract insights","Sensitive data (API keys, user information) may be logged; requires careful configuration to avoid security issues"],"requires":["Python 3.9+","Logging configuration (file, database, or external service)","Optional: external tracing service (e.g., Langsmith, Arize)"],"input_types":["agent execution events","logging configuration"],"output_types":["execution traces","log files","performance metrics"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_8","uri":"capability://search.retrieval.web.search.integration.for.real.time.information.retrieval","name":"web search integration for real-time information retrieval","description":"Integrates web search capabilities into agents, allowing them to retrieve current information from the internet when needed. Agents can decide when to search, formulate queries, and incorporate results into responses. Supports multiple search providers (Google, DuckDuckGo, Bing) and handles result parsing, ranking, and deduplication automatically.","intents":["I want my agent to search the web for current information instead of relying only on training data","I need agents to answer questions about recent events or real-time data","I want to build a research agent that gathers information from multiple web sources"],"best_for":["developers building agents that need access to current information","teams creating research or news aggregation agents","applications requiring real-time data that changes frequently"],"limitations":["Web search results are noisy; no built-in filtering for relevance or quality","Search latency adds significant overhead to agent response time (typically 1-5 seconds per search)","Search providers have rate limits and may block automated queries; requires careful management","Results may contain outdated, incorrect, or malicious information; agents may hallucinate based on poor search results"],"requires":["Python 3.9+","API key for search provider (Google Custom Search, DuckDuckGo, Bing, etc.)","Network connectivity for web requests"],"input_types":["search queries","search configuration (provider, result count, language)"],"output_types":["search results (title, URL, snippet)","parsed web content","aggregated information"],"categories":["search-retrieval","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__cap_9","uri":"capability://automation.workflow.asynchronous.agent.execution.with.concurrent.tool.calls","name":"asynchronous agent execution with concurrent tool calls","description":"Supports asynchronous execution of agents and concurrent invocation of multiple tools in parallel. Agents can await results from multiple tools simultaneously, reducing latency for workflows with independent tool calls. Built on Python's asyncio, enabling integration with async frameworks (FastAPI, aiohttp, etc.) without blocking.","intents":["I want my agent to call multiple tools in parallel to reduce total execution time","I need to integrate agents into async web frameworks without blocking","I want to build high-concurrency agent systems that handle many simultaneous requests"],"best_for":["developers building high-performance agent APIs","teams integrating agents into async web frameworks (FastAPI, Starlette)","applications requiring low-latency agent responses with multiple tool calls"],"limitations":["Async code is more complex to debug; errors in concurrent tasks may be silent if not properly handled","Tool execution order becomes non-deterministic; some workflows may require sequential tool calls","Concurrent tool calls may hit rate limits or resource constraints; requires careful management of concurrency levels","Not all tools support async execution; blocking tools will stall concurrent execution"],"requires":["Python 3.9+","Async-compatible tools and dependencies","Understanding of Python asyncio patterns"],"input_types":["async functions","concurrent tool definitions"],"output_types":["aggregated results from concurrent tools","execution timing data"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"phidata__headline","uri":"capability://memory.knowledge.ai.agent.framework.with.memory.and.knowledge.integration","name":"ai agent framework with memory and knowledge integration","description":"Phidata is a comprehensive framework for building AI agents that can utilize memory, knowledge, and tools, enabling sophisticated function calling and structured outputs for various AI models.","intents":["best AI agent framework","AI framework for building intelligent agents","top frameworks for RAG and memory integration","AI agents with multi-agent capabilities","best tools for developing AI agents"],"best_for":["developers building AI agents","teams needing multi-agent systems"],"limitations":[],"requires":[],"input_types":[],"output_types":[],"categories":["memory-knowledge","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":58,"verified":false,"data_access_risk":"high","permissions":["Python 3.9+","API keys for at least one provider (OpenAI, Anthropic, Google, etc.) OR local Ollama instance","phidata package installed via pip","LLM provider that supports function calling (OpenAI, Anthropic, Google, etc.)","Pydantic for type hints (optional but recommended for complex schemas)","LLM capable of multi-step reasoning (GPT-4, Claude, etc.)","Understanding of prompt engineering best practices","Document files in supported formats (PDF, markdown, text, HTML)","Optional: PDF parsing libraries (PyPDF2, pdfplumber)","Python 3.8+"],"failure_modes":["Provider-specific features (e.g., vision capabilities, function calling schemas) may not be fully normalized across all providers","Response latency varies significantly by provider; no built-in load balancing or failover between providers","Some advanced parameters (e.g., Anthropic's thinking budget) may not map cleanly to the unified interface","Schema generation from Python functions may fail for complex types (nested generics, custom classes); requires explicit Pydantic models for reliability","No built-in retry logic for failed tool calls; requires manual implementation of error handling and recovery","Async tool execution requires careful management of event loops; blocking tools will stall the entire agent","Chain-of-thought adds latency; LLM must generate reasoning steps before final answer","Reasoning quality depends on LLM capability; weaker models may produce poor reasoning","No automatic verification of reasoning correctness; agents may show confident but incorrect reasoning","Reasoning steps may be verbose or confusing to users; requires careful prompt engineering","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7,"quality":0.9,"ecosystem":0.49999999999999994,"match_graph":0.25,"freshness":0.52,"weights":{"adoption":0.3,"quality":0.2,"ecosystem":0.15,"match_graph":0.23,"freshness":0.12}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-06-17T09:51:05.295Z","last_scraped_at":null,"last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=phidata","compare_url":"https://unfragile.ai/compare?artifact=phidata"}},"signature":"31tEF1Lz11T4J+xsWQ0mt7KGmyZVs4OAPqdutJfteM42m8Aoxn53koY7R9LtR/+0B7Mn6wZnJlo9++IU3fxYDg==","signedAt":"2026-06-20T06:37:51.950Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/phidata","artifact":"https://unfragile.ai/phidata","verify":"https://unfragile.ai/api/v1/verify?slug=phidata","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}