Dakera retriever

Dakera is a self-hosted memory server that provides persistent, decay-weighted vector recall. The DakeraRetriever lets an agent fetch the most relevant documents for a query from a Dakera namespace, using Dakera’s text-query API (embeddings are generated server-side), so it can be used as a retrieval-augmented context source.

Dakera runs beside your app; the canonical way to run it is the dakera-deploy docker-compose stack (server + object store, default port 3000).

Installation

pip install "agent-squad[dakera]"

Configuration

DakeraRetrieverOptions accepts:

| Option | Default | Description | | ----------- | ------------------------- | ------------------------------------------------------------------------------ | | namespace | (required) | The Dakera namespace to query. | | api_key | DAKERA_API_KEY env | Dakera API key (a dk-... token). | | url | DAKERA_URL env, else http://localhost:3000 | Base URL of the Dakera server. | | top_k | 10 | Maximum number of results to return. | | filter | None | Optional Dakera metadata filter applied to the query. |

Using the Dakera retriever

You can add a DakeraRetriever to a BedrockLLMAgent (or any agent that accepts a retriever), so the LLM generates responses grounded in the documents retrieved from your Dakera namespace:

Python

from agent_squad.agents import BedrockLLMAgent, BedrockLLMAgentOptions
from agent_squad.retrievers import DakeraRetriever, DakeraRetrieverOptions

orchestrator.add_agent(
    BedrockLLMAgent(BedrockLLMAgentOptions(
        name="My personal agent",
        description="Answers questions using context retrieved from Dakera.",
        streaming=True,
        inference_config={
            "temperature": 0.1,
        },
        retriever=DakeraRetriever(DakeraRetrieverOptions(
            namespace="my-docs",
            top_k=5,
        )),
    ))
)

TypeScript

import { BedrockLLMAgent, DakeraRetriever } from "agent-squad";

orchestrator.addAgent(
  new BedrockLLMAgent({
    name: "My personal agent",
    description: "Answers questions using context retrieved from Dakera.",
    streaming: true,
    inferenceConfig: {
      temperature: 0.1,
    },
    retriever: new DakeraRetriever({
      namespace: "my-docs",
      topK: 5,
    }),
  })
);

You can also use the retriever on its own:

import asyncio
from agent_squad.retrievers import DakeraRetriever, DakeraRetrieverOptions

retriever = DakeraRetriever(DakeraRetrieverOptions(namespace="my-docs", top_k=5))

# Raw results (list of TextSearchResult with .id / .score / .text / .metadata)
results = asyncio.run(retriever.retrieve("How many languages are spoken worldwide?"))

# Or a single combined context string, ready to pass to an LLM
context = asyncio.run(
    retriever.retrieve_and_combine_results("How many languages are spoken worldwide?")
)

Set DAKERA_API_KEY (and optionally DAKERA_URL) in your environment, or pass api_key/url explicitly in the options. DakeraRetriever is retrieval-only, so retrieve_and_generate is not supported — use retrieve or retrieve_and_combine_results and let your agent’s LLM generate the response.