AI providers
AI providers are organization-wide. Spinners and Call Analyzer reference them by name; they do not store their own copies of API keys.
| Kind | Used by |
|---|---|
| LLM | AI agent steps, PCA scoring / GenAI tags |
| STT | PCA transcription |
| Storage | PCA recording download / storage |
Add an LLM
Section titled “Add an LLM”- Open Integrations → AI providers.
- Open the LLM tab.
- Add a provider, choose a model, and fill in credentials.
- Save — credentials are encrypted in the secured vault.
Need two models with the same API key? Create two integrations (e.g. “OpenAI — Haiku” and “OpenAI — GPT-4o”) with different model selections.
Supported LLM providers
Section titled “Supported LLM providers”| Provider | Credentials / config |
|---|---|
| OpenAI | API key, model |
| Anthropic | API key, model |
| Groq | API key, model |
| Azure OpenAI | API key, endpoint URL, deployment name (acts as model), optional API version |
| Google Gemini | Google AI Studio API key, model |
| Gemini (Vertex AI) | GCP project ID, region, service account JSON key, model |
| Ollama (local) | Base URL, model name, optional API key |
| Mock | No credentials — for pipeline tests without spending tokens |
Add a transcriber (STT)
Section titled “Add a transcriber (STT)”Same panel, Transcriber tab. Pick the provider and model here — Call Analyzer setup only chooses which integration to use and the call language.
For regional Indian languages, prefer OpenAI Whisper or a dedicated Deepgram integration with the right model. Analyzer setup shows a compatibility badge when the language may not match the transcriber.
Add storage
Section titled “Add storage”Storage tab — configure bucket/container, optional path prefix, and credentials. Call Analyzer projects reference the integration; they do not re-enter bucket names.
Use in a spinner
Section titled “Use in a spinner”On an AI agent step, select the LLM integration (model is already on that integration).
Use in Call Analyzer
Section titled “Use in Call Analyzer”On Analyzer setup, pick storage, transcriber, and LLM integrations only. Language, corrections, and scoring stay on the project.
Security
Section titled “Security”Keys and service account JSON are stored encrypted in the Secured Vault. They are not shown again in full after save. Credential fields show a shield icon — hover for details.