MuleSoft AI Chain (MAC) Vectors + Einstein + Chroma

Simple Mule application making use of the MAC (MuleSoft AI Chain) project to read documents from a given URL and parse them to a JSON String. Using Einstein as language model and Chroma as the vectors database.

Similar repos

1 - Salesforce Einstein (Model)

• Make sure you have access to Einstein in Salesforce. Retrieve your organization's URL. For example, https://mulesoft-dev.develop.lightning.force.com/

• Go to Setup > App Manager > New Connected App > Create a Connected App.

  • Connected App Name: MuleSoft app
  • Contact Email: your email
  • Enable OAuth Settings: ✅
  • Callback URL: https://login.salesforce.com/services/oauth2/callback
  • Selected OAuth Scopes: Access Einstein GPT services (einstein_gpt_api)
  • Enable Client Credentials Flow: ✅

• Press Save.

• In the created Connected App detail page, click on Manage Consumer Details.

• Copy the consumer key and consumer secret. These are your client id/secret respectively.

2 - Chroma (Vector DB)

• Make sure you have Docker Desktop installed.

• Follow the steps here or run:

  docker pull chromadb/chroma
  docker run -p 8000:8000 chromadb/chroma

3 - API Specification

• Go to Design Center in your Anypoint Platform account.
• Create a new RAML 1.0 specification and copy/paste the contents of this file or import it directly.
• Publish to Exchange.

4 - Implement the Mule app

Note

You can compare your code to this

• Install the MAC Vectors module with Maven by following this doc.

• Add the following Global Elements to your XML file:

  <ms-vectors:config name="MuleSoft_Vectors_Connector_Config" doc:name="MuleSoft Vectors Connector Config" doc:id="14c92cd7-e040-4634-9bb8-04a26539f756">
      <ms-vectors:embedding-model-service>
          <ms-vectors:einstein salesforceOrgUrl="your-sf-org-url" clientId="sf-connectedapp-clientid" clientSecret="sf-connectedapp-clientsecret" />
      </ms-vectors:embedding-model-service>
      <ms-vectors:vector-store>
          <ms-vectors:chroma url="localhost:8000" />
      </ms-vectors:vector-store>
  </ms-vectors:config>
  <file:config name="File_Config" doc:name="File Config" doc:id="4333d39a-0253-4d3b-86f1-c9f168b0920e" />
  <http:request-config name="HTTP_Request_configuration" doc:name="HTTP Request configuration" doc:id="1ac8c8d2-c885-43a8-a30e-96d7a21e789a">
      <http:request-connection host="#[vars.uri.host]" />
  </http:request-config>
  <http:request-config name="HTTPS_Request_configuration" doc:name="HTTP Request configuration" doc:id="88a2acf4-445a-4dbb-afaf-c7cb132cb9ea">
      <http:request-connection host="#[vars.uri.host]" protocol="HTTPS" />
  </http:request-config>

• Replace the Salesforce credentials with your own.

• Make sure Chroma is indeed running on port 8000 or modify it from the XML.

• Head to the last flow and update it with the next:

  <flow name="post:\reader\url:text\plain:mac-file-reader-config">
      <ee:transform doc:name="Transform Message" doc:id="533db0b0-a059-4170-82a1-c1ab13172465">
          <ee:message />
          <ee:variables>
              <ee:set-variable variableName="uri"><![CDATA[%dw 2.0
  output application/json
  import substringAfter, substringBefore, substringBeforeLast from dw::core::Strings
  var host = (payload substringAfter "://") substringBefore "/"
  var path = ((payload substringAfter host) substringBeforeLast "/") ++ "/"
  ---
  {
      protocol: lower(payload substringBefore ":"),
      host: host,
      path: path,
      fileName: (payload substringAfter path)
  }]]></ee:set-variable>
              <ee:set-variable variableName="workingDir"><![CDATA[%dw 2.0
  output application/json
  ---
  (mule.home default "") ++ "/apps/" ++ (app.name default "")]]></ee:set-variable>
          </ee:variables>
      </ee:transform>
      <choice doc:name="Choice" doc:id="d3934daf-b5c0-4546-94a9-c9dcd8e2ef89">
          <when expression="#[&quot;https&quot; == vars.uri.protocol]">
              <http:request method="GET" doc:name="Request" doc:id="9a407153-2627-4cf9-a776-4078ec1b237a" config-ref="HTTPS_Request_configuration" path="#[vars.uri.path ++ vars.uri.fileName]" />
          </when>
          <otherwise>
              <http:request method="GET" doc:name="Request" doc:id="42240506-b8d5-4417-8e95-842106ea7d53" config-ref="HTTP_Request_configuration" path="#[vars.uri.path ++ vars.uri.fileName]" />
          </otherwise>
      </choice>
      <file:write doc:name="Write" doc:id="ec1fead7-0437-4088-8ffb-fb8cf00845c1" config-ref="File_Config" path="#[vars.workingDir ++ vars.uri.fileName]" />
      <ms-vectors:document-parser doc:name="Document parser" doc:id="0bd74ed0-6448-4ff2-bb5d-4ecc99361534" config-ref="MuleSoft_Vectors_Connector_Config" fileType="any" contextPath="#[vars.uri.fileName]">
          <ms-vectors:storage>
              <ms-vectors:local workingDirecotry="#[vars.workingDir]" />
          </ms-vectors:storage>
      </ms-vectors:document-parser>
      <ee:transform doc:name="Transform Message" doc:id="2add1957-ebe5-468d-b081-3d6d789f074e">
          <ee:message>
              <ee:set-payload><![CDATA[%dw 2.0
  output text/plain
  ---
  payload.text]]></ee:set-payload>
          </ee:message>
      </ee:transform>
  </flow>