> For the complete documentation index, see [llms.txt](https://docs.scanbylens.xyz/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.scanbylens.xyz/the-engine/engine.md).

# Overview

Capture, normalize, connect, flag, narrate, verify.

The engine is the intelligence of LensAI. The language model only narrates what the engine has already measured.

```
capture ──▶ factor state store ──▶ mechanism graph ──▶ divergence engine
   │          (normalize against own history,                │
   │           regime-tag, one clock)                        ▼
   └──────────────────────────────▶ tool layer ◀──── narration + verification
                                        │                    │
                                        ▼                    ▼
                                 outside agents      briefs, calls, the record
```

| Stage                                                      | Job                                                                                                            | Detail                                                             |
| ---------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ |
| [**Capture**](/the-engine/capture.md)                      | Record every article and market observation, immutably, with the time LensAI learned it.                       | Polled feeds every few minutes to daily; liquidations pushed live. |
| [**Factor state store**](/the-engine/factor-store.md)      | Turn every raw stream into a percentile against its own history, tag its regime, join everything on one clock. | Hourly anchors; 90-day, 365-day and full-history windows.          |
| [**Mechanism graph**](/the-engine/mechanism-graph.md)      | Record the documented channels through which one factor moves another.                                         | Curated, versioned, append-only.                                   |
| [**Divergence engine**](/the-engine/divergence.md)         | Flag extreme states, broken relationships and structural signatures.                                           | Pure arithmetic.                                                   |
| [**Narration and verification**](/the-engine/narration.md) | Write the brief and propose the call, then verify every claim before release.                                  | One structured generation; fail-closed gates.                      |
| [**Tool layer**](/the-engine/tool-layer.md)                | Expose the stores as read-only, point-in-time tools for the desk and for outside agents.                       | JSON-RPC over HTTP and stdio.                                      |

## Design rules that run through every stage

* **Point-in-time everywhere.** Every read takes an "as of" moment and cannot see anything captured after it. Lookahead bias is structurally impossible, not merely discouraged.
* **Append-only.** Raw observations, the mechanism graph, briefs and calls are never updated or deleted. Database triggers reject mutation, and tests prove it.
* **Fail closed.** Missing data produces an honest gap, a withheld value or a no call. It never produces an estimate.
* **Model-agnostic.** The engine does not depend on any one language model. Models can be swapped or added without changing what makes LensAI valuable.
* **Tested.** The engine carries a test suite that runs on every change, including red-team suites for the non-advisory guardrail and the call policy.

## Infrastructure

The engine runs as one always-on service: capture daemon, live feeds, the hourly derive cycle (normalize, regimes, divergence, shared brief, watches) and the HTTP API, in a single process. The corpus is stored in SQLite with portable SQL and a migration path to Postgres when a second service needs it, and is set up for continuous off-site replication of the database, because the corpus is the one asset that cannot be rebuilt. The website and desk are a Next.js application served separately.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.scanbylens.xyz/the-engine/engine.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.