# AI use cases for Lossless Memory (aru-labs/lossless-memory)

Repo: [aru-labs/lossless-memory](https://github.com/aru-labs/lossless-memory) | Use cases: 4 | Page: https://dedupedai.com/use-cases/aru-labs-lossless-memory/

A local, file-based long-term memory layer for a personal AI that keeps every conversation line verbatim, timestamps each record, and searches by time before words.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Keep every line of a personal AI's chats on your machine: https://dedupedai.com/use-cases/aru-labs-lossless-memory/keep-full-ai-chat-history-locally/
- Pull the exact lines from a past evening or day: https://dedupedai.com/use-cases/aru-labs-lossless-memory/recall-what-was-said-on-a-given-day/
- Import Claude Code transcripts into a searchable memory: https://dedupedai.com/use-cases/aru-labs-lossless-memory/import-claude-code-transcripts-into-memory/
- Keep a long-running assistant on topic after context compaction: https://dedupedai.com/use-cases/aru-labs-lossless-memory/keep-assistant-oriented-after-context-compaction/

## How to keep every line of a personal AI's chats on your machine with Lossless Memory

Repo: [aru-labs/lossless-memory](https://github.com/aru-labs/lossless-memory) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Store raw conversation logs verbatim in local JSONL files and build search indexes from them.

**Who it is for:** A developer who talks to a personal AI every day and wants nothing summarized, so the exact wording and timestamps survive across sessions.

**The problem:** Summaries and similarity search drop the exact words and the time they were said, so the assistant later forgets what was actually decided.

**The outcome:** A local folder of per-day JSONL logs plus SQLite indexes, all rebuildable from the logs, with every record stamped with its time.

### Steps

1. `git clone https://github.com/aru-labs/lossless-memory`
2. `cd lossless-memory`
3. `pip install -e .`
4. `cp config.example.json config.json`
5. Edit config.json to set your names, paths, and timezone (an IANA name such as America/New_York).
6. Follow examples/quickstart.md to ingest the sample conversation and build the indexes.

### Prompt to paste into your agent

```text
Using the repo aru-labs/lossless-memory, set up the memory layer on <target machine> following examples/quickstart.md. Expected output: the sample conversation ingested into per-day JSONL logs and the exact and vector indexes built under data_dir. Verify that the raw log files exist and that the 7-field records include a ts value.
```

### Caveat

The README says it has only run for one person on one machine, so there is no multi-user setup to rely on.

Risk, credentials: The logs store conversation text verbatim, so any passwords or keys pasted into chats will be kept in plain files on disk.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/aru-labs-lossless-memory/keep-full-ai-chat-history-locally/

## How to pull the exact lines from a past evening or day with Lossless Memory

Repo: [aru-labs/lossless-memory](https://github.com/aru-labs/lossless-memory) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Ask for a time phrase like yesterday evening and get the verbatim lines from that range in order.

**Who it is for:** A person who wants to check what an AI assistant agreed to on a specific day, and needs the exact words rather than a paraphrase.

**The problem:** Similarity search returns loosely related chunks from weeks ago, not the lines from the time you remember.

**The outcome:** Results restricted to the requested time range, in chronological order, unsummarized, with their timestamps.

### Steps

1. `git clone https://github.com/aru-labs/lossless-memory`
2. `cd lossless-memory`
3. `pip install -e .`
4. `cp config.example.json config.json`
5. Set timezone in config.json so that yesterday and similar phrases resolve to your local day.
6. Ingest the logs and run a time-scoped query as shown in examples/quickstart.md, using a phrase such as last Tuesday night or 2026-07-19.

### Prompt to paste into your agent

```text
Using the repo aru-labs/lossless-memory, run a time-scoped recall for the phrase <time phrase> against the logs in <target data directory> following examples/quickstart.md. Expected output: verbatim log lines from that range in chronological order with timestamps. Verify that every returned line falls inside the parsed date range.
```

### Caveat

Relative time phrases are parsed only in English and Japanese, so other languages need absolute dates such as 2026-07-19.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/aru-labs-lossless-memory/recall-what-was-said-on-a-given-day/

## How to import Claude Code transcripts into a searchable memory with Lossless Memory

Repo: [aru-labs/lossless-memory](https://github.com/aru-labs/lossless-memory) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Point the importer at Claude Code transcript files and build a time-indexed memory from them.

**Who it is for:** A developer who uses Claude Code daily and wants its past sessions searchable by time without writing a custom importer.

**The problem:** Past coding sessions live in transcript files that cannot be searched by when they happened or by the exact words used.

**The outcome:** Transcript lines converted into timestamped records in the lossless log and indexed for time-scoped recall.

### Steps

1. `git clone https://github.com/aru-labs/lossless-memory`
2. `cd lossless-memory`
3. `pip install -e .`
4. `cp config.example.json config.json`
5. In config.json, set ingest_format to "claude_code" and point the input path at your transcript files.
6. Run the ingest and index steps described in examples/quickstart.md, then confirm the records appear in the per-day logs.

### Prompt to paste into your agent

```text
Using the repo aru-labs/lossless-memory, configure ingest_format as claude_code and import the transcript files at <transcript path> following examples/quickstart.md. Expected output: per-day JSONL logs containing the imported transcript lines with ts values. Verify that the record count matches the number of transcript messages you expect.
```

### Caveat

The README calls the Claude Code importer the path with the most mileage, but another app would need its own importer.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/aru-labs-lossless-memory/import-claude-code-transcripts-into-memory/

## How to keep a long-running assistant on topic after context compaction with Lossless Memory

Repo: [aru-labs/lossless-memory](https://github.com/aru-labs/lossless-memory) | Difficulty: Advanced | Time: 1 hour | Basis: From the README

Maintain a short timestamped topic index that is injected into the model's context every turn.

**Who it is for:** A person running a long-lived AI assistant whose context window gets compacted, so it loses track of the current thread.

**The problem:** After compaction or a new session, the assistant has no clear record of which subject the conversation is on.

**The outcome:** A small LLL topic index the model reads each turn, with priority and completion marks set by the human.

### Steps

1. `git clone https://github.com/aru-labs/lossless-memory`
2. `cd lossless-memory`
3. `pip install -e .`
4. Read docs/lll.md to learn the human and AI division of labor for topic markers.
5. Write short timestamped topic marker lines whenever the conversation moves to a new subject.
6. Inject the rendered LLL index into the model context at the start of each turn.

### Prompt to paste into your agent

```text
Using the repo aru-labs/lossless-memory, read docs/lll.md and set up an LLL topic index for <assistant name>. Expected output: a short list of timestamped topic markers that can be injected each turn. Verify that the human-set priority and completion marks are not edited by the model.
```

### Caveat

The README describes LLL as a design and gives no ready-made injection command, so you must wire it into your own assistant.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/aru-labs-lossless-memory/keep-assistant-oriented-after-context-compaction/