> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bensevern.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# GoldenGraph overview

> Build an own-your-KG knowledge graph from text: LLM extraction, goldenmatch entity resolution, and a durable bi-temporal store.

GoldenGraph builds an **own-your-KG** knowledge graph from text:

**text → LLM extraction → [GoldenMatch](/docs/goldenmatch/overview) entity resolution → a
durable, bi-temporal store.**

Entity resolution is the differentiator. Duplicate surface forms across documents collapse
into one durable entity — the step most GraphRAG frameworks do badly, because they treat
resolution as string matching rather than as the entity-resolution problem it actually is.

```python theme={null}
from goldengraph import ingest, OpenAIClient
from goldengraph_native import _native as gg

store = gg.PyStore()                    # durable bi-temporal store
llm = OpenAIClient()                    # or any object with .complete(prompt) -> str
ingest("Acme Inc, founded by ...", store, at=1, llm=llm)

snapshot = store.snapshot()             # canonical JSON; persist + reopen later
view = store.as_of(valid_t=1, tx_t=1)   # bi-temporal slice -> queryable graph
```

## The stages

| Stage                | Does                                                             |
| -------------------- | ---------------------------------------------------------------- |
| `extract(text, llm)` | Text to typed entities and relationships, via the LLM.           |
| `resolve(...)`       | Collapses duplicate surface forms using goldenmatch's ER engine. |
| `ingest(...)`        | Runs the whole path and writes into the store.                   |

## Bi-temporal by construction

The store records both **valid time** (when a fact was true in the world) and
**transaction time** (when the graph learned it). `as_of(valid_t, tx_t)` returns the graph
as it was believed at a point in time — which is what makes a KG auditable rather than
merely current. A correction does not erase what the graph previously asserted; it adds a
new transaction over the same valid-time span.

## Relationship to goldenmatch-kg

GoldenGraph is a **complete KG builder**. [GoldenMatch KG](/docs/goldenmatch-kg/overview) is the
opposite shape: it drops goldenmatch's resolution into *someone else's* KG pipeline
(neo4j-graphrag, LlamaIndex, Graphiti). Use GoldenGraph when you own the pipeline; use
goldenmatch-kg when you are adopting one.