===============
LIBRARY RULES
===============
From library maintainers:
- Eidentic is a TypeScript SDK published on npm as `eidentic` (umbrella) and `@eidentic/*` (32 packages). Always use the TypeScript API and ESM imports.
- Construct an agent with `new Agent({ id, model, store })`; wrap any AI SDK provider via `new AIModel(provider(modelId))`.
- Use `@eidentic/libsql` (LibsqlStore, pure-JS, no native addon) for Next.js/serverless; `@eidentic/sqlite` (SqliteStore) for Node/Bun.
- `agent.query(input, { sessionId })` returns an async generator of StreamEvents; text deltas are `ev.type === 'stream.delta'` carrying `ev.delta.text`; the terminal event is `ev.type === 'result'`.
- Memory: `new Memory({ store, vector, embedder, graph })` — four tiers (lexical+vector recall, self-editing blocks, temporal knowledge graph, consolidation).
- Next.js: `export const POST = withEidentic(agent)` from `@eidentic/nextjs`. React: `useEidenticStream(endpoint)` from `@eidentic/react`. Server: `createServer({ agents })` + `serveNode(app, { port })` from `@eidentic/server`.
- The umbrella package `eidentic` re-exports core+types+model+sqlite+memory; install adapters (`@eidentic/server`, `@eidentic/workflow`, etc.) separately as needed.

### Install and Build Eidentic Examples

Source: https://github.com/eidentic/eidentic/blob/main/README.md

Installs dependencies, builds packages, and runs a 'hello' command with a mock model. No API key is needed for this step.

```bash
pnpm install
pnpm -r build
pnpm --filter eidentic-examples hello
```

--------------------------------

### Install Dependencies and Run Development Server

Source: https://github.com/eidentic/eidentic/blob/main/packages/create-eidentic/README.md

After scaffolding, navigate to the project directory, install dependencies using pnpm, set your API key environment variable, and start the development server.

```bash
cd my-agent
pnpm install
# Set your API key, e.g.:
export ANTHROPIC_API_KEY=sk-ant-...
pnpm dev
```

--------------------------------

### Install Eidentic and AI SDK

Source: https://github.com/eidentic/eidentic/blob/main/README.md

Install the necessary packages for Eidentic and the Anthropic AI SDK. This is the initial setup step for using Eidentic in your application.

```bash
npm install eidentic ai @ai-sdk/anthropic
```

--------------------------------

### Basic Usage with Node.js

Source: https://github.com/eidentic/eidentic/blob/main/packages/server/README.md

Set up and run the Eidentic server in a Node.js environment. This example demonstrates agent creation, authentication setup, and starting the server on port 3000. The server exposes endpoints for querying agents, checking run status, and listing workflows.

```typescript
import { createServer, serveNode, ApiKeyAuth } from "@eidentic/server";
import { Agent, AIModel, SqliteStore } from "eidentic";
import { anthropic } from "@ai-sdk/anthropic";

const agent = new Agent({
  id: "support",
  model: new AIModel(anthropic("claude-sonnet-4-5")),
  store: new SqliteStore("./eidentic.sqlite"),
});

const app = createServer({
  agents: { support: agent },
  auth: ApiKeyAuth({ "key_live_abc123": { userId: "u-1" } }),
});

// Node.js
await serveNode(app, { port: 3000 });
// POST /v1/agents/support/query  →  NDJSON stream
// GET  /v1/runs/:id/status       →  async run status
// GET  /v1/workflows             →  workflow run list

```

--------------------------------

### Run Context Compaction Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:compaction' example to test token-budget-triggered window compaction that preserves failure evidence and atomic tool pairs.

```bash
pnpm -C examples hello:compaction
```

--------------------------------

### Install @eidentic/prompts

Source: https://github.com/eidentic/eidentic/blob/main/packages/prompts/README.md

Install the library using pnpm.

```bash
pnpm add @eidentic/prompts
```

--------------------------------

### Run MCP Host Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:mcp' example to test consuming external MCP servers' tools as first-class Eidentic tools.

```bash
pnpm -C examples hello:mcp
```

--------------------------------

### Run Reasoning Strategies Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:strategies' example to test composable reasoning strategies like 'reflection' (using a different model as a critic) and 'planAndExecute'.

```bash
pnpm -C examples hello:strategies
```

--------------------------------

### Run Token Streaming Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:stream' example to test the agent's ability to stream model output token-by-token.

```bash
pnpm -C examples hello:stream
```

--------------------------------

### Run Prompt Skills Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:skill' example to test the agent's ability to search and use skills defined in SKILL.md files.

```bash
pnpm -C examples hello:skill
```

--------------------------------

### Run Any Eidentic Feature Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute any feature example by specifying its name after 'hello:'.

```bash
# run any feature:

```

```bash
pnpm -C examples hello:<name>
```

--------------------------------

### Install @eidentic/libsql

Source: https://github.com/eidentic/eidentic/blob/main/packages/libsql/README.md

Install the package using pnpm.

```bash
pnpm add @eidentic/libsql
```

--------------------------------

### Initialize Eidentic Agent with PostgreSQL Store

Source: https://github.com/eidentic/eidentic/blob/main/packages/postgres/README.md

Set up a PostgreSQL connection pool and initialize an Eidentic agent using the PostgresStore. This example demonstrates how to connect to a database and start an agent interaction.

```typescript
import { PostgresStore } from "@eidentic/postgres";
import pg from "pg";
import { Agent, AIModel } from "eidentic";
import { anthropic } from "@ai-sdk/anthropic";

const pool = new pg.Pool({ connectionString: process.env.DATABASE_URL });
const store = await PostgresStore.create({ client: pool });

const agent = new Agent({
  id: "support",
  model: new AIModel(anthropic("claude-sonnet-4-5")),
  store,
});

for await (const ev of agent.query("Hello", { sessionId: "s-1" })) {
  if (ev.type === "stream.delta") process.stdout.write(ev.delta.text);
}
```

--------------------------------

### Run MCP Server Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:mcp-server' example to test exposing Eidentic tools/agents as an MCP server compatible with any MCP client.

```bash
pnpm -C examples hello:mcp-server
```

--------------------------------

### Run Schema Evolution Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:schema-evolution' example to test event upcasting for replaying old logs on new code and embedding re-indexing.

```bash
pnpm -C examples hello:schema-evolution
```

--------------------------------

### Install @eidentic/skills

Source: https://github.com/eidentic/eidentic/blob/main/packages/skills/README.md

Install the package using pnpm.

```bash
pnpm add @eidentic/skills
```

--------------------------------

### Install PostgreSQL Adapter and pg

Source: https://github.com/eidentic/eidentic/blob/main/packages/postgres/README.md

Install the @eidentic/postgres package and the 'pg' library using pnpm.

```bash
pnpm add @eidentic/postgres pg
```

--------------------------------

### Run Sleep-Time Consolidation Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:consolidation' example to test background distillation of episodes into durable facts, grounded and without invention.

```bash
pnpm -C examples hello:consolidation
```

--------------------------------

### Install @eidentic/bench

Source: https://github.com/eidentic/eidentic/blob/main/packages/bench/README.md

Install the benchmark package using pnpm.

```bash
pnpm add @eidentic/bench
```

--------------------------------

### Install @eidentic/workflow

Source: https://github.com/eidentic/eidentic/blob/main/packages/workflow/README.md

Install the workflow package using pnpm.

```bash
pnpm add @eidentic/workflow
```

--------------------------------

### Install @eidentic/server

Source: https://github.com/eidentic/eidentic/blob/main/packages/server/README.md

Install the @eidentic/server package using pnpm.

```bash
pnpm add @eidentic/server
```

--------------------------------

### Run Temporal Knowledge Graph Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:graph' example to test the temporal knowledge graph, which stores facts with validity over time and handles contradictions.

```bash
pnpm -C examples hello:graph
```

--------------------------------

### Install @eidentic/better-auth and better-auth

Source: https://github.com/eidentic/eidentic/blob/main/packages/better-auth/README.md

Install the necessary packages for the better-auth session adapter.

```bash
pnpm add @eidentic/better-auth better-auth
```

--------------------------------

### Run Executable Skills Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:executable-skill' example to test gated executable skills, allowed tools, ed25519 signing, and quarantine mechanisms.

```bash
pnpm -C examples hello:executable-skill
```

--------------------------------

### Install @eidentic/tools

Source: https://github.com/eidentic/eidentic/blob/main/packages/tools/README.md

Install the @eidentic/tools package using pnpm.

```bash
pnpm add @eidentic/tools
```

--------------------------------

### Run Durable Execution Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:durable' example to test checkpoint/resume functionality and exactly-once tool dispatch that survives crashes.

```bash
pnpm -C examples hello:durable
```

--------------------------------

### Run Skill Self-Evolution Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:skill-evolution' example to test a skill improving its own playbook, using unit tests as the reflection signal (off by default).

```bash
pnpm -C examples hello:skill-evolution
```

--------------------------------

### Install @eidentic/convex

Source: https://github.com/eidentic/eidentic/blob/main/packages/convex/README.md

Install the Eidentic Convex adapter and its peer dependency Convex.

```bash
npm i @eidentic/convex convex
```

--------------------------------

### Install Qdrant Adapter

Source: https://github.com/eidentic/eidentic/blob/main/packages/qdrant/README.md

Install the Qdrant vector store adapter for Eidentic using pnpm.

```bash
pnpm add @eidentic/qdrant
```

--------------------------------

### Run Multi-Agent Delegation Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:multi-agent' example to test agent delegation using 'spawn_agent' with context isolation and a shared tree budget.

```bash
pnpm -C examples hello:multi-agent
```

--------------------------------

### Install Dependencies and Build Project

Source: https://github.com/eidentic/eidentic/blob/main/CONTRIBUTING.md

Installs dependencies, compiles native addons, builds all packages, runs the full test suite, and typechecks every package.

```bash
pnpm install          # installs deps and compiles native addons
pnpm -r build         # build all packages (emits dist/)
npx vitest run        # run the full test suite
pnpm -r typecheck     # typecheck every package
```

--------------------------------

### Run Hosted Embedder Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:hosted-embedding' example to test using hosted embeddings with your own API key across any AI SDK provider.

```bash
pnpm -C examples hello:hosted-embedding
```

--------------------------------

### Run Agent-to-Agent (A2A) Protocol Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:a2a' example to test the Agent-to-Agent protocol for exposing and consuming remote agents.

```bash
pnpm -C examples hello:a2a
```

--------------------------------

### Install @eidentic/sqlite

Source: https://github.com/eidentic/eidentic/blob/main/packages/sqlite/README.md

Install the SQLite store adapter package using pnpm.

```bash
pnpm add @eidentic/sqlite
```

--------------------------------

### Run Cancellation Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:cancellation' example to test cooperative aborts, resulting in an 'aborted' status, checkpointing, and child teardown.

```bash
pnpm -C examples hello:cancellation
```

--------------------------------

### Install HuggingFace Hub Library

Source: https://github.com/eidentic/eidentic/blob/main/packages/bench/BASELINES.md

Installs the necessary Python library for interacting with the HuggingFace Hub.

```bash
pip install huggingface_hub
```

--------------------------------

### Run Memory Maintenance Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:memory-maintenance' example to test fact staleness, archival deduplication/merging, and single-flight consolidation.

```bash
pnpm -C examples hello:memory-maintenance
```

--------------------------------

### Run Right to Erasure (GDPR) Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:erasure' example to test the hard deletion of all data associated with a scope across stores, vectors, and graphs.

```bash
pnpm -C examples hello:erasure
```

--------------------------------

### Install @eidentic/studio with pnpm

Source: https://github.com/eidentic/eidentic/blob/main/packages/studio/README.md

Install the studio package using pnpm. This command adds the necessary dependencies to your project.

```bash
pnpm add @eidentic/studio
```

--------------------------------

### Run Passive Extraction and Scopes Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:memory-completion' example to test auto fact extraction without an LLM and organization/shared memory scopes.

```bash
pnpm -C examples hello:memory-completion
```

--------------------------------

### Run Self-Editing Memory Blocks Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:self-editing' example to test the agent's ability to edit its own always-in-context memory during reasoning.

```bash
pnpm -C examples hello:self-editing
```

--------------------------------

### Install pgvector Adapter

Source: https://github.com/eidentic/eidentic/blob/main/packages/pgvector/README.md

Install the pgvector adapter for Eidentic. Choose between a standard PostgreSQL setup or an in-process PGlite setup.

```bash
pnpm add @eidentic/pgvector pg
# or with pglite (in-process, no server needed):
pnpm add @eidentic/pgvector @electric-sql/pglite
```

--------------------------------

### Quickstart Eidentic Agent

Source: https://github.com/eidentic/eidentic/blob/main/packages/umbrella/README.md

Initialize an Eidentic agent with a model and a durable store, then query it for information. This example uses the Anthropic Claude Sonnet model and a local SQLite store.

```typescript
import { Agent, AIModel, SqliteStore } from "eidentic";
import { anthropic } from "@ai-sdk/anthropic";

const agent = new Agent({
  id: "support",
  model: new AIModel(anthropic("claude-sonnet-4-5")),
  store: new SqliteStore("./eidentic.sqlite"),
});

for await (const ev of agent.query("What did we decide last week?", { sessionId: "u-42" })) {
  if (ev.type === "stream.delta") process.stdout.write(ev.delta.text);
  if (ev.type === "result") console.log("\nDone:", ev.subtype);
}
```

--------------------------------

### Run Built-in Tools Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:tools' example to test the agent's built-in tools, including file, bash, and web tools with security guards.

```bash
pnpm -C examples hello:tools
```

--------------------------------

### Standalone Node.js Eidentic Server

Source: https://github.com/eidentic/eidentic/blob/main/docs/DEPLOYMENT.md

Sets up a standalone Eidentic agent server using `@eidentic/server`. This example configures an agent with a specific model and store, sets up API key authentication, and starts the server on a specified port.

```typescript
// server.ts
import { Agent, AIModel, SqliteStore } from "eidentic";
import { createServer, serveNode, ApiKeyAuth } from "@eidentic/server";
import { anthropic } from "@ai-sdk/anthropic";

const store = new SqliteStore("./eidentic.sqlite");

const agent = new Agent({
  id: "support",
  model: new AIModel(anthropic("claude-sonnet-4-5")),
  store,
  instructions: "You are a helpful support assistant.",
});

const app = createServer({
  agents: { support: agent },
  auth: ApiKeyAuth({ [process.env.API_KEY!]: { userId: "service" } }),
});

await serveNode(app, { port: Number(process.env.PORT ?? 3000) });
console.log(`Listening on :${process.env.PORT ?? 3000}`);

```

--------------------------------

### Quick Start: Registering and Deploying Prompts

Source: https://github.com/eidentic/eidentic/blob/main/packages/prompts/README.md

Demonstrates registering new prompt versions, deploying a version as 'stable', and retrieving the stable prompt. Identical prompt bodies are de-duplicated.

```typescript
import { createPromptRegistry, filePromptStore, renderPrompt } from "@eidentic/prompts";

const registry = createPromptRegistry(filePromptStore("./data/prompts.json"));

// Register versions (identical body = no-op)
const v1 = await registry.register("support-agent-system", "You are a helpful assistant.");
const v2 = await registry.register("support-agent-system", "You are a concise, helpful assistant.");

// Deploy v2 as stable
await registry.tag("support-agent-system", 2, "stable");

// Resolve the stable prompt per request
const prompt = await registry.get("support-agent-system", "stable");
```

--------------------------------

### Integrate and Serve Studio with Custom Agents

Source: https://github.com/eidentic/eidentic/blob/main/packages/studio/README.md

Integrate the Eidentic Studio with your custom agents and serve it on a specified port. This TypeScript example shows how to create a studio instance, associate it with agents, and start the development server. Authentication is optional and can be configured separately.

```typescript
import { createStudio, serveNode } from "@eidentic/studio";

const studio = createStudio({
  agents: { support: myAgent },
  // auth: ApiKeyAuth({ "dev-key": { userId: "dev" } }), // optional
});

await serveNode(studio, { port: 4000 });
// Open http://localhost:4000 in your browser
```

--------------------------------

### Install @eidentic/core Package

Source: https://github.com/eidentic/eidentic/blob/main/packages/core/README.md

Install the core Eidentic package using pnpm. Ensure you have the 'ai' peer dependency installed.

```bash
pnpm add @eidentic/core
```

--------------------------------

### Run Lazy Tool Discovery Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:lazy-tools' example to test lazy tool discovery using 'search_tools' and 'load_tool' for smaller manifests with large toolsets.

```bash
pnpm -C examples hello:lazy-tools
```

--------------------------------

### Install @eidentic/langfuse

Source: https://github.com/eidentic/eidentic/blob/main/packages/langfuse/README.md

Install the package using npm.

```bash
npm install @eidentic/langfuse
```

--------------------------------

### Run Stateful ReAct Agent Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello' example to test the core Stateful ReAct agent, which uses a reason → tool → observe loop.

```bash
pnpm -C examples hello
```

--------------------------------

### Install @eidentic/types

Source: https://github.com/eidentic/eidentic/blob/main/packages/types/README.md

Install the @eidentic/types package using pnpm.

```bash
pnpm add @eidentic/types
```

--------------------------------

### Run Remote Vector Stores Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:remote-vector' example to test adapters for remote vector stores like pgvector, Qdrant, and Pinecone, including in-memory fakes.

```bash
pnpm -C examples hello:remote-vector
```

--------------------------------

### Run Real Model Agent Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:real' example to test the agent against a real AI provider using AI SDK v6. Requires an API key.

```bash
pnpm -C examples hello:real
```

--------------------------------

### Install @eidentic/mcp

Source: https://github.com/eidentic/eidentic/blob/main/packages/mcp/README.md

Install the MCP package using pnpm.

```bash
pnpm add @eidentic/mcp
```

--------------------------------

### Install @eidentic/eval Package

Source: https://github.com/eidentic/eidentic/blob/main/packages/eval/README.md

Install the @eidentic/eval package using pnpm.

```bash
pnpm add @eidentic/eval
```

--------------------------------

### Install @eidentic/a2a Package

Source: https://github.com/eidentic/eidentic/blob/main/packages/a2a/README.md

Install the A2A package using pnpm.

```bash
pnpm add @eidentic/a2a
```

--------------------------------

### Install RAG Package

Source: https://github.com/eidentic/eidentic/blob/main/packages/rag/README.md

Install the core @eidentic/rag package using pnpm.

```bash
pnpm add @eidentic/rag
```

--------------------------------

### Run Embedded Usage Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:embedded' example to test dropping the agent into a plain Node.js HTTP server without requiring a dedicated server package.

```bash
pnpm -C examples hello:embedded
```

--------------------------------

### Install @eidentic/cli

Source: https://github.com/eidentic/eidentic/blob/main/packages/cli/README.md

Install the eidentic command-line tool globally using npm.

```bash
npm install -g @eidentic/cli
# or use via the umbrella (eidentic ships the CLI)
npm install eidentic
```

--------------------------------

### Initialize LibsqlStore and Agent

Source: https://github.com/eidentic/eidentic/blob/main/packages/libsql/README.md

Demonstrates initializing the LibsqlStore for both local file-based and remote Turso connections, and then creating an Eidentic Agent with it. Ensure necessary environment variables are set for Turso.

```typescript
import { LibsqlStore } from "@eidentic/libsql";
import { Agent, AIModel } from "eidentic";
import { anthropic } from "@ai-sdk/anthropic";

// Local file (development)
const store = new LibsqlStore("file:eidentic.db");

// Turso remote (production)
const store = new LibsqlStore({
  url: process.env.LIBSQL_URL!,
  authToken: process.env.LIBSQL_AUTH_TOKEN,
});

const agent = new Agent({
  id: "support",
  model: new AIModel(anthropic("claude-sonnet-4-5")),
  store,
});

for await (const ev of agent.query("Hello", { sessionId: "s-1" })) {
  if (ev.type === "stream.delta") process.stdout.write(ev.delta.text);
}
```

--------------------------------

### Run Lexical Recall and Memory Blocks Example

Source: https://github.com/eidentic/eidentic/blob/main/docs/TESTING.md

Execute the 'hello:memory' example to test lexical recall and always-in-context memory blocks, which require no external infrastructure.

```bash
pnpm -C examples hello:memory
```

--------------------------------

### Install better-sqlite3 Peer Dependency

Source: https://github.com/eidentic/eidentic/blob/main/packages/sqlite/README.md

Install the better-sqlite3 peer dependency separately as required by the adapter.

```bash
pnpm add better-sqlite3
```

--------------------------------

### Run End-to-End Demo

Source: https://github.com/eidentic/eidentic/blob/main/CONTRIBUTING.md

Executes the end-to-end 'hello' demo, which uses a scripted mock model and does not require an API key.

```bash
pnpm --filter eidentic-examples hello
```