diff --git a/marketing/src/pages/index.astro b/marketing/src/pages/index.astro index ecdf7ef..b0c97fe 100644 --- a/marketing/src/pages/index.astro +++ b/marketing/src/pages/index.astro @@ -26,15 +26,18 @@ import SEO from "../components/SEO.astro";
Persistent memory for AI agents
-

+

deja

-

+

Agents forget everything between runs. deja fixes that.

-

- An open-source memory layer that stores what agents learn and recalls it when relevant. - Available as three packages — pick the runtime that fits your stack. +

+ A memory layer. Agents call learn() to store what they know + and recall() to get it back. +

+

+ Three packages — local, edge, hosted — same idea, pick your runtime.

@@ -61,167 +64,161 @@ import SEO from "../components/SEO.astro";
- -
-
- What is deja? -
-
-
-
-

- Every time an AI agent runs, it starts from scratch. It doesn't remember what worked last time, - what failed, or what another agent already figured out. You end up re-explaining the same context, - re-discovering the same workarounds, and watching agents make the same mistakes. -

-

- deja is a memory layer. Agents call learn() to store - what they discover and recall() to retrieve what's relevant - before they act. Memories have confidence scores, get deduplicated automatically, and decay when they go stale. -

-

- It works across agents, across runs, across tools. CI learns something → your chat agent already knows it. - One agent figures out a workaround → every agent benefits. -

-
-
-
- - -
+ +
Three memory systems
-

- Same concepts, different runtimes. Pick the one that matches your stack, or use them together. +

+ Same concepts, different runtimes. Pick the one that matches your stack.

-
- -
-
-
- deja-local -
-
- In-process vector memory for Bun -
+ +
+ + + +
-
-
- SEARCH - Vector similarity (all-MiniLM-L6-v2 via ONNX) -
-
- STORAGE - SQLite file on disk -
-
- LATENCY - Zero — runs in your process -
-
- DEPS - Bun runtime + HuggingFace transformers + +
+ +
+
+
+
+ In-process vector memory for Bun +
+
+
+ SEARCH + Vector similarity (all-MiniLM-L6-v2 via ONNX) +
+
+ STORAGE + SQLite file on disk +
+
+ LATENCY + Zero — runs in your process +
+
+ DEPS + Bun runtime + HuggingFace transformers +
+
+

+ Best for: CLI tools, local agents, scripts, development. +

-
- -
import { DejaLocal } from 'deja-local' +
+
import { DejaLocal } from 'deja-local' const mem = new DejaLocal('./memories.db') await mem.remember("node 20 breaks esbuild") const hits = await mem.recall("deploying")
- -

- Best for: CLI tools, local agents, scripts, development. -

+
+
-
-
-
- deja-edge -
-
- Edge memory for Cloudflare Durable Objects -
- -
-
- SEARCH - FTS5 full-text search (BM25 + Porter stemming) -
-
- STORAGE - Durable Object SQLite -
-
- LATENCY - Zero — runs inside the DO -
-
- DEPS - None — zero external dependencies + -
-
-
- deja (hosted) -
-
- Full-featured memory service on Cloudflare -
- -
-
- SEARCH - Semantic vector search (Vectorize + Workers AI) -
-
- STORAGE - DO SQLite + Vectorize index -
-
- LATENCY - Network hop — deployed to your CF account -
-
- DEPS - Cloudflare Vectorize + Workers AI bindings +
-
+
@@ -317,119 +314,6 @@ mem.remember("node 20 breaks esbui - -
-
- Try it — hosted demo -
-
-
-
- -
-
- Teach -
-
-
- - -
-
- - -
- -
-
-
- - -
-
- Recall -
-
-
- - -
- -
-
-
- Memories will appear here... -
-
-
-
- - -
-
- POST - /learn - -
-
-
-
- - -
-
Honest answers
-
-
-
-

"Why three packages?"

-

Different agents run in different places. A CLI tool needs in-process memory with zero latency. An edge worker needs memory inside its Durable Object. A multi-agent team needs shared memory over HTTP. Same core idea, three runtimes.

-
-
-
-

"Why not just use a database?"

-

You could. But then you're building semantic search, deduplication, confidence scoring, conflict resolution, and cleanup yourself. deja is that layer.

-
-
-
-

"Vendor lock-in?"

-

deja-local runs on Bun with SQLite — no cloud required. deja-edge runs inside any Cloudflare DO. The hosted version runs on your own CF account. MIT license, fork it if we disappear.

-
-
-
-

"Which one should I use?"

-

Building a local CLI tool or script? deja-local. Building on Cloudflare Workers? deja-edge. Need shared memory across multiple agents or MCP? deja (hosted).

-
-
-
-
remember("node 20 breaks esbui