egma-ai/egma / skills/integrate-egma
integrate-egma
VerifiedIntegrate Egma into a repository containing voice agent code. Use when the developer asks to integrate voice agent testing, monitoring or both
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.0 KB |
| agents/openai.yaml | config | 355 B |
| references/livekit-agent-connection-guide.md | markdown | 3.6 KB |
| references/pipecat-agent-connection-guide.md | markdown | 12.1 KB |
| references/retell-agent-connection-guide.md | markdown | 3.0 KB |
| references/setup-monitoring.md | markdown | 6.3 KB |
| references/setup-simulation-testing.md | markdown | 4.7 KB |
references/setup-simulation-testing.md
View on GitHub# Guide to setup simulation testing for voice agents with egma platform
The broad order of things that need to be setup before we can run successful tests is as follows:
1. There needs to be an egma folder in this repo and it should be in sync with the platform.
2. Create an agent in egma platform and a connection to it so that it can be reached by a simulated persona for testing purposes.
3. Author tests and how they need to be evaluated.
4. Now, we can run the test suite against the connected agent.
1. Setup the egma folder and sync state
- Egma allows authoring voice agent tests locally as simple files so that tests can be maintained alongside the code. All egma tests should live inside a `egma/` folder in the repo.
- Check if an `egma/` folder currently exists in this repo holding tests and config.
- If it does not exist - run the `egma init` command to bootstrap it. This will create generate a `egma/config.yaml` file and an empty `egma/tests/` folder to store test suites. Once authored, the state can must be synced with the egma platform using `egma pull` & `egma push` commands. Run `egma pull` right now to pre-emptlively fetch if there are any agents or tests authored already via the platform UI.
- If an `egma/` folder already exists, that means someone has already setup egma in this repo and you just need to run the pull command to fetch the latest state from remote.
- Unless explicitly asked to use a self hosted version of the egma platform - default to egma cloud platform hosted at https://app.egma.ai/
2. Connect the agent in this repo with egma platform
- The egma platform needs to reach the agent in this repo in order to run simulation tests with it. There are different connection strategies based on the agent platform and modality of connection. Example - retell agent with chat modality (i.e simulation testing a retell voice agent but without voice - just testing the behavior of the main LLM in a "STT-LLM-TTS" cascaded voice agent setup)
- First off, identify which platform the agent in this repo built on. Currently egma supports connecting with agents built on livekit agents (python/ js), pipecat (python) & retell.
- Based on the platform - refer to the platform specific connection guides. Follow the guides and come back here for following the next step.
- For connecting a livekit agent - you must look at [guide to connect a livekit agent](./livekit-agent-connection-guide.md) which is a part of this skill's references
- For connecting a pipecat bot (pipecat-ai in the dependencies, a `bot(runner_args)` entry point, often a `pcc-deploy.toml`) - you must look at [guide to connect a pipecat agent](./pipecat-agent-connection-guide.md) which is a part of this skill's references
- For connecting a retell agent - you must look at [guide to connect a retell agent](./retell-agent-connection-guide.md) which is a part of this skill's references
3. Author tests
- If this is the first time egma tests are being written in this repository, create at max four tests in one suite covering 4 most common scenrios the voice agent here encounters.
- Refer to the `/write-voice-agent-tests` skill. This will teach you how to create a test suite and write good tests that actually move the needle in terms of trusting the voice agent's behavior. Egma tests are just simple markdown files. If this skill is not installed, you MUST first install it using `npx skills add egma-ai/egma --yes`. It contains important instructions regarding test creation, tool mocking and dynamically passing session data to simulation environment.
- For pipecat, run the first suite on the path the developer chose in the pipecat guide: this machine through `egma agent dev`, Pipecat Cloud, or the team's own servers. Ask before any change to production.
- For livekit, when the SDK changes are still only on local, temporarily give the local worker a unique agent name and use it on both egma connections so that cloud deployed workers cannot pick up the simulations and error on the simulation run. Keep it running until the suite finishes. After the first run finishes/ before stopping and handing off to the developer, restore the original name in the worker and connections, stop the local worker, and tell the developer to deploy the SDK changes before starting new runs and running more simulation tests.
4. Run the tests
- Once there is a suite you wanna run use `egma run --help` to look at how to run a suite against a particular agent and one of its connections. If the agent has multiple connections, default to using the voice connection for the run. Once the run has started, immediately give the developer the link to UI so they can see the run progressing visually. However, your job is only complete once the suite has finished running!
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |