# AI use cases for Apowerb (apowerb/apowerb) Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Use cases: 5 | Page: https://dedupedai.com/use-cases/apowerb-apowerb/ apowerb is an open-source platform for building, running and governing AI agents in production, self-hosted with Docker Compose or Helm. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a self-hosted AI agent platform with Docker Compose: https://dedupedai.com/use-cases/apowerb-apowerb/self-host-agent-platform-in-minutes/ - Triage new Gmail or Outlook messages with an agent: https://dedupedai.com/use-cases/apowerb-apowerb/triage-inbox-with-email-triggered-agent/ - Send a scheduled morning report from a SQL database: https://dedupedai.com/use-cases/apowerb-apowerb/scheduled-morning-report-from-sql/ - Build a document assistant over your own files and URLs: https://dedupedai.com/use-cases/apowerb-apowerb/answer-questions-from-company-documents/ - Create CRM deals and update ERP records with an agent: https://dedupedai.com/use-cases/apowerb-apowerb/automate-crm-and-erp-updates-unattended/ ## How to run a self-hosted AI agent platform with Docker Compose with Apowerb Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README Start the web interface, API and database on one machine with three commands. **Who it is for:** An IT or operations person who needs a private, self-hosted place to build AI agents and does not want to set up a database by hand. **The problem:** Setting up an agent platform usually means wiring a database, secrets and services before anything works. **The outcome:** A running stack with the web interface on port 3000, the API on port 8000, and a PostgreSQL database included. ### Steps 1. Clone the hosting repo: `git clone https://github.com/apowerb/apowerb-hosting.git && cd apowerb-hosting` 2. Create the config file and random secrets: `cp .env.example .env && ./scripts/generate-secrets.sh` 3. Start the stack: `docker compose -f docker-compose/docker-compose.yml --env-file .env up -d` 4. Open http://localhost:3000 and add your model key in the interface, or set DEFAULT_LLM_MODEL and DEFAULT_LLM_API_KEY in .env. ### Prompt to paste into your agent ```text Using the apowerb-hosting repo at https://github.com/apowerb/apowerb-hosting, start the Docker Compose stack on this machine and report the web interface URL. Verify that http://localhost:3000 loads and that the API answers on port 8000. The model list stays empty until a model key is added, so state whether one is set. ``` ### Caveat Agents cannot answer until a model key is added, because the README says this part cannot be generated for you. Risk, credentials: The model key and the generated secrets in .env grant access to paid model usage and must be kept out of version control. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/apowerb-apowerb/self-host-agent-platform-in-minutes/ ## How to triage new Gmail or Outlook messages with an agent with Apowerb Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Difficulty: Advanced | Time: 1 hour | Basis: From the README A new email starts an agent that classifies it, checks the CRM and drafts a reply. **Who it is for:** A customer support or sales lead whose inbox fills faster than the team can sort it, and who wants drafts ready before a person reads each message. **The problem:** Each incoming message needs reading, sorting, a customer lookup and a first reply, all done by hand. **The outcome:** Each new message gets a classification, a CRM lookup and a draft reply, with a session trace for review. ### Steps 1. Start the stack from the quick start: `docker compose -f docker-compose/docker-compose.yml --env-file .env up -d` 2. In the web interface, define an agent with instructions, a model and the tools it needs. 3. Connect the Gmail or Outlook account that will trigger the agent, following the Webhooks section of the docs. (inferred, check the README) 4. Add a CRM tool such as HubSpot or Pipedrive so the agent can look up the sender. 5. Send a test message and open its session trace to check the classification and the draft. ### Prompt to paste into your agent ```text Using the apowerb repo at https://github.com/apowerb/apowerb, build an inbox triage agent that is triggered by new messages in . Expected output: for each message, a category, a CRM lookup result and a draft reply. Verify the session trace records each run before trusting the drafts. ``` ### Caveat Gmail and Outlook triggers depend on webhook callbacks reaching PUBLIC_BASE_URL, which must be a public address for a real mailbox to reach the server. Risk, credentials: The agent needs access to a mailbox and a CRM, so the account grants should be limited to what triage needs. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/apowerb-apowerb/triage-inbox-with-email-triggered-agent/ ## How to send a scheduled morning report from a SQL database with Apowerb Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Difficulty: Advanced | Time: 1 hour | Basis: From the README A scheduled agent asks the database in plain language and mails the summary. **Who it is for:** A manager who wants the same daily numbers in the inbox each morning without writing queries. **The problem:** Daily figures live in a database, and someone has to write the query, make a chart and send it every day. **The outcome:** An agent runs on a schedule, turns a plain-language question into SQL, and mails a summary with a chart. ### Steps 1. Start the stack: `docker compose -f docker-compose/docker-compose.yml --env-file .env up -d` 2. Add your SQL database as a tool so the agent can run Text-to-SQL against it. 3. Define an agent that asks the question in plain language, builds a chart and writes the summary. 4. Set the cron schedule for the run, which the README says is driven by an external orchestrator. (inferred, check the README) 5. Run the agent once by hand and check the session trace before the first scheduled send. ### Prompt to paste into your agent ```text Using the apowerb repo at https://github.com/apowerb/apowerb, create a scheduled agent that answers against with Text-to-SQL and mails a summary with a chart. Expected output: one email with the summary and chart. Verify the generated SQL in the session trace matches the question. ``` ### Caveat The README does not say how the external orchestrator is set up, so the scheduling step needs to be confirmed in the docs before relying on it. Risk, credentials: The database login used by the agent should be read-only, since generated SQL runs against live data. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/apowerb-apowerb/scheduled-morning-report-from-sql/ ## How to build a document assistant over your own files and URLs with Apowerb Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Index files, URLs or S3 content into a knowledge base and let an agent answer from it. **Who it is for:** A team lead or knowledge manager who wants staff to ask questions about contracts or procedures without searching folders. **The problem:** Answers sit in scattered contracts, procedures and web pages, and finding them takes time. **The outcome:** An agent answers questions using your indexed documents, and you can check each answer against its source file. ### Steps 1. Start the stack: `docker compose -f docker-compose/docker-compose.yml --env-file .env up -d` 2. Create a knowledge base and add your files, URLs, databases or S3 content through RAG. 3. Define an agent in the web interface and attach the knowledge base to it. 4. Ask the agent three questions whose answers you already know from the source documents. ### Prompt to paste into your agent ```text Using the apowerb repo at https://github.com/apowerb/apowerb, index into a knowledge base and attach it to an agent. Expected output: answers to five test questions. Verify each answer against the original document before the agent is shared. ``` ### Caveat The README does not describe how answers are cited, so checking answers against source files is a manual step. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/apowerb-apowerb/answer-questions-from-company-documents/ ## How to create CRM deals and update ERP records with an agent with Apowerb Repo: [apowerb/apowerb](https://github.com/apowerb/apowerb) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Let an agent create deals or update Odoo records from a request or on its own. **Who it is for:** An operations manager who repeats the same data entry in HubSpot, Pipedrive or Odoo and wants it handled by an agent with an audit trail. **The problem:** Creating deals and updating ERP records by hand is slow and easy to get wrong. **The outcome:** An agent makes the CRM or ERP changes from a conversation or unattended, with every run logged in a session trace. ### Steps 1. Start the stack: `docker compose -f docker-compose/docker-compose.yml --env-file .env up -d` 2. Connect the HubSpot, Pipedrive or Odoo tools that the agent needs. 3. Define the agent's instructions so it only creates deals or updates the record types you list. 4. Set a per-user token quota for the agent's user so usage stays bounded. 5. Run one request in the chat, then review the session trace before enabling unattended runs. ### Prompt to paste into your agent ```text Using the apowerb repo at https://github.com/apowerb/apowerb, build an agent that creates a deal in from and updates the matching Odoo record. Expected output: one new deal and one updated record, both listed in the session trace. Verify each change in the CRM and Odoo before running unattended. ``` ### Caveat Business-system tools act on live records, and the README does not describe a dry-run mode, so test on a copy or a sandbox account first. Risk, credentials: The CRM and Odoo accounts the agent uses can change live business records, so grant only the permissions the agent needs. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/apowerb-apowerb/automate-crm-and-erp-updates-unattended/