TotalRecall™ — The AI That Never Forgets

Skill Details

Support Skills Help Center Partners
Help Center › Skills › TotalRecall
v2.3 — Free Update

TotalRecall™

By ProductiveBot · Version 2.3.0 · Last updated May 5, 2026

The AI that never forgets. TotalRecall gives your ProductiveBot persistent memory across every conversation, every channel, and every document — and now searches all of it with one command. Works on every messaging platform you connect.

Setup
Pre-installed
Privacy
100% local
Extra cost
None

The Problem

Without TotalRecall, your ProductiveBot starts fresh every conversation. It does not remember your name, your preferences, your past decisions, or what you talked about yesterday. Every session is a blank slate.

This is not a bug. It is how most AI assistants work by default. The model processes your message, responds, and moves on. There is no built-in system to carry information forward.

TotalRecall changes that. It adds a persistent memory layer that runs entirely on your Mac Mini and remembers what matters across every conversation, channel, and document you bring into your bot's world.

How It Works

TotalRecall searches everything your bot has access to and brings the relevant context back when you need it. It indexes your conversations, documents, Slack history, and notes automatically — no setup beyond install.

In v2.3, one unified command (search-all.sh) searches every available source in parallel and tags each result by where it came from. Your bot tells you exactly where it found something — your March 27 conversation, a PDF from last week, or a Slack thread from two months ago.

Unified Search

One command searches memory, sessions, documents, and Slack at once. Results are tagged with source so you always know where information came from.

Every Conversation, Every Platform

Every message that goes through your bot — Slack, Telegram, Signal, web chat — is automatically searchable. Platform-specific integrations like Slack add deeper history on top.

Full Transcript Search

Search every past conversation on your Mac Mini. Recall what you discussed weeks or months ago, even sessions that were never saved to memory files.

Documents and PDFs

Shared files, workspace docs, and PDFs are searchable too. Your bot can find that contract, that proposal, or that meeting note when you ask for it.

Platform Detection

Your bot tells you exactly what it can search on this unit — what's installed, what's synced, when it was last updated. No guessing.

Completely Private

All memory and search runs on your Mac Mini. No cloud, no telemetry, no third party. Your conversations stay yours.

Platform Support

TotalRecall searches across every source your unit has access to. Sources are auto-detected — anything not installed is gracefully skipped, with no errors and no setup friction.

Source Status Notes
Session Transcripts ✓ Always available Covers every conversation, every platform — Slack, Telegram, Signal, web
Memory Files ✓ Always available Curated long-term memory and daily notes
Documents and PDFs ✓ Always available Shared files, workspace docs, full PDF text extraction
Slack (slacrawl) ✓ When connected Full history, all channels and DMs, auto-synced every 30 min
Telegram Coming soon Sessions cover Telegram conversations in the meantime
WhatsApp Coming soon Sessions cover WhatsApp conversations in the meantime
Discord Coming soon Sessions cover Discord conversations in the meantime
Semantic search Future Available when a local model is installed (Mac Mini bundle or ProductiveBot+)
One key thing: every conversation that goes through your bot is captured automatically — no matter which platform it came in on. Crawlers like slacrawl add deeper platform-specific history on top, but the core recall layer works on every unit, every channel, out of the box.

Works Without a Local Model. Even Better With One.

TotalRecall is fully functional without any local AI model. Everything you have read so far works on every ProductiveBot — no model required. If a local model is installed on your unit, TotalRecall uses it for smarter semantic search. Two modes are supported by design.

Concurrent Mode

Local model loaded and running. Your bot can chat with it AND TotalRecall can use it for semantic search at the same time. This is the default setup on units that ship with a local model.

Offline Mode

You unload the local model to free up memory for something else. TotalRecall keeps working, just with keyword search instead of semantic. When you reload the model, semantic search comes back automatically.

You do not need to switch modes manually. Unload or reload your local model the way you normally would — TotalRecall figures it out.

See It In Action

Here is what a conversation looks like with TotalRecall installed:

Week 1
"Remember that our fiscal year starts in April, not January."
"Got it. I've noted that your fiscal year starts in April."
3 Weeks Later, New Conversation
"Can you put together a Q1 revenue summary?"
"Sure. Since your fiscal year starts in April, Q1 would be April through June. Let me pull that together..."
2 Months Later — Cross-Source Recall
"What did we decide about the Henderson account?"
"I found three places where Henderson came up:
[SESSION] March 12 — pricing discussion with Sam
[SLACK] #sales — final terms posted April 3
[DOC] Henderson_proposal.pdf — signed contract
Want me to pull any of these up?"
Without TotalRecall, the bot in the second conversation would assume January and the third would have no memory of Henderson at all. With TotalRecall, your business context compounds.

Memory File Structure

TotalRecall uses simple markdown files that you and your bot both read and write:

workspace/
  ├── MEMORY.md       # Long-term: key decisions, preferences
  └── memory/
      ├── 2026-03-15.md  # Daily notes
      ├── 2026-03-16.md
      └── 2026-03-17.md

MEMORY.md is for curated, long-term memories: preferences, key decisions, lessons learned. Think of it as your bot's cheat sheet.

Daily files are raw notes from each day. Your bot writes these automatically as it works. Over time, the important stuff gets distilled into MEMORY.md.

What It Does Not Do

Transparency matters. Here is what TotalRecall is not:

It does not record your conversations verbatim from outside your bot. The session layer captures conversations your bot was part of. It does not silently log what you say in other apps.

It does not send data anywhere. There is no cloud sync, no analytics, no telemetry. Memory and search both run on your Mac Mini.

It is not AI training. Your memories are never used to train any AI model. They are only used to give your bot context in your conversations.

Common Questions

How do I install TotalRecall?
TotalRecall ships pre-installed on every ProductiveBot unit. If you have an older unit without it, visit the Skill Store and click Install on the TotalRecall card.
Does it slow down my bot?
No. Searches typically finish in under 10 seconds, even across very large histories. The unified search runs all sources in parallel, so adding more sources does not significantly slow things down.
What platforms does it search?
Every conversation your bot is part of is searchable through session history — Slack, Telegram, Signal, web chat, anywhere. Slack also has a deeper integration that searches your full workspace history. Telegram, WhatsApp, and Discord deeper integrations are on the roadmap.
Does it work without a local AI model?
Yes. TotalRecall is fully functional without any local model. Search works on every unit, model or no model. If a local model is installed, TotalRecall uses it for smarter semantic search. If you unload the model later to free up RAM, TotalRecall switches to keyword search and keeps working.
Can I read and edit the memory files myself?
Yes. They are plain markdown files. You can open them in any text editor, add notes, correct mistakes, or delete things you do not want remembered. Changes are picked up automatically.
How much storage does it use?
Memory files are plain text and stay tiny — typically under a few megabytes after months of use. Slack history (if connected) takes more, around 100 MB per 50,000 messages. Even after a year of heavy use, the total footprint fits comfortably on any Mac Mini.
What if I want my bot to forget something?
Open the memory file and delete the line. Or tell your bot "forget that my favorite color is blue" and it will remove it from MEMORY.md. Changes take effect immediately.
Is TotalRecall included with every ProductiveBot?
Yes. TotalRecall ships pre-installed on every new ProductiveBot. Existing units can install it for free from the Skill Store. It is a free update for all ProductiveBot owners.
Does it work with ChatGPT, Claude, or other providers?
Yes. TotalRecall is independent of which AI model your bot uses. It works with Anthropic (Claude), OpenAI (ChatGPT), and any other supported provider. The memory and search system is the same regardless of provider.

Update Log

v2.3.0 May 5, 2026 CURRENT

Unified search across all sources in one command. Platform detection — your bot tells you exactly what it can search. Source-tagged results [MEMORY] [SESSION] [DOC] [SLACK]. Two-mode architecture for local model users (concurrent and offline). Foundation for semantic search via Ollama embed models.

v2.0.0 April 26, 2026

Crawl Army integration — Slack, Notion, Discord, X/Twitter, WhatsApp offline mirrors. slacrawl auto-sync every 30 min via launchd agent. 8-step Deep Recall protocol.

v1.8.0 March 27, 2026

Full session transcript search across all past conversations on disk. Document and PDF search (shared files, workspace docs). Google Docs integration. Recency-weighted results. 7-step Deep Recall protocol. Mandatory session summarization.

v1.6.0 March 24, 2026

Session and channel history search. Deep Recall protocol. Session summarization. Improved installer with native module rebuild.

v1.0.0 March 11, 2026

Initial release — local memory file search, BM25 keyword matching.

Need help? Talk to Scout

Scout can help you install TotalRecall, troubleshoot memory issues, or answer any questions.

You can also email us at support@productivebot.ai

ProductiveBot.ai · Free for all ProductiveBot owners