Audifact Beta
PERSONAL & HOBBY

Your personal agents,
with a memory you can trust

Weekend project? Homework helper? Notes search bot? Wrap your chain in one line and see exactly what it did — no enterprise setup, no credit card, and your data stays yours.

Try the 2-minute example Get a free API key (optional)

Runs great in demo mode — no API key needed to play locally

SOUND FAMILIAR?

Cool agent… but what did it
actually read and say?

🧠

Learning by tinkering

You're experimenting with LangChain or LlamaIndex for the first time. A simple audit trail helps you understand what each step did — great for students and self-teaching.

🔒

Privacy on personal data

Your notes, journal, or home docs shouldn't leak into random logs. Default hashes-only mode means proofs without exposing your private text.

🎯

"Why did it answer that?"

Personal RAG and research agents are fun until they hallucinate. A linked decision → outcome pair lets you replay what happened in plain language.

GET STARTED IN 60 SECONDS

Installation

Simplest path — LangChain
pip install "audifact-sdk[langchain]"
Three lines: import tracer → .wrap(your_chain) → invoke. That's it.
Or try LlamaIndex
pip install "audifact-sdk[llamaindex]"
Same idea with AudifactLlamaIndexTracer — great for personal doc Q&A
Demo mode: run examples with no AUDIFACT_API_KEY — perfect for learning on your laptop. Add a key later when you want events in the explorer.
SUPER SIMPLE EXAMPLES

Projects you can build this weekend

PERSONAL RAG

Ask questions about your notes

from audifact.langchain import AudifactLangChainTracer

# Your chain: load notes → retrieve → answer
chain = build_notes_qa_chain()  # however you built it

tracer = AudifactLangChainTracer(
    agent_name="my-notes-bot",
    # no api_key → demo mode, totally fine!
)
answer = tracer.wrap(chain).invoke(
    "What did I write about the Japan trip?"
)

Home lab, study notes, recipe collection, journal entries — anything you'd grep manually but wish an agent could summarize.

example script →
RESEARCH BUDDY

A tiny agent that cites its sources

from audifact.langchain import AudifactLangChainTracer

agent = build_research_agent(tools=[web_search, summarize])

tracer = AudifactLangChainTracer(
    agent_name="weekend-researcher",
    include_context=True,  # see tool steps when debugging
)
tracer.wrap(agent).invoke({
    "input": "Best beginner keyboards for programming in 2026?"
})

Great for hobby deep-dives, coursework, or comparing gear — you'll see each tool call and the final answer linked together.

LlamaIndex version →

Home automation, personal finance helpers, fan-fiction research — same pattern, same one-line wrap.

WHAT YOU'LL SEE

Friendly event pairs — not scary audit logs

Each run creates a decision (what you asked) and an outcome (what the agent answered). Readable labels, no jargon required.

💬
Your question
chain_execution
The prompt you sent — hash by default. Explorer shows public_decision_label when you set one.
✨
The answer
chain_response
Short summary on public proof: public_outcome_label: Chain Response: …
🛠️
Agent at work
agent_execution
Marks the start of an agent run — handy when it calls multiple tools.
📎
Final result
agent_response
Final reply + optional tool trail. Click decision → outcome in the Explorer.
WHY PEOPLE LOVE IT FOR PERSONAL PROJECTS

Serious enough to learn from.
Light enough to enjoy.

✓
Literally one line to add

tracer.wrap(chain) — no refactors, no new architecture. Keep hacking on your project.

✓
Zero overhead to start

Demo mode works offline. Free API key when you're ready — still no credit card.

✓
Your data stays yours

Hashes by default. You choose include_context=True only when you want full text stored — your call.

✓
Grows if you do

Same tools the startups and enterprises use. Start personal, scale later — or don't. Totally fine.

Go build something fun.
Bring receipts.

Wrap your chain, run the example, peek in the explorer. You'll get it in one coffee break.

See the examples Get a free API key

Scaling up later? Check out startups and enterprise pages.