Inference Conversation History (Beta)

Telnyx Inference now supports Conversation History, the ability to persist, embed, index, and semantically search call transcripts across your Telnyx communications. Enable it on a SIP connection and search via API, with no external vector database, no pipeline glue, and no data leaving the private backbone.

What's new

• Conversation persistence on SIP connections: Enable conversation_persistence on any SIP connection (IP, credential, or FQDN). When a call completes, Telnyx automatically embeds the transcript, chunks it, and indexes it.
• Semantic search via API: Query across all persisted conversations with GET /v2/ai/conversation_histories. Results come back chunked and relevance-scored, ready to use as RAG context for your agent.
• Filter and scope support: Narrow results by region, record type, and additional filters to pinpoint the conversations that matter.
• On-network processing: Embedding and indexing run on Telnyx Inference infrastructure. The call happens on our SIP network, the transcript gets embedded on our GPU fleet, and search runs through the same API. One system, one vendor, one contract.
• 30-day indexed retention included: No extra config required. Extended retention and additional storage options are available for longer recall windows.

Why it matters

• Teams building AI agents on real-time communications typically stitch together four separate vendors: a telephony provider, a transcription service, a vector database, and an inference API. Conversation History collapses that into a single SIP connection flag.
• The Frankenstack integration tax is real. Every vendor boundary is re-integration engineering at every API change, and each hop adds 30-80ms of latency. On-network embedding eliminates those boundaries.
• Search results are chunked and relevance-scored out of the box, making them immediately usable as RAG context. No vector DB setup, no embedding pipeline, no schema management.
• Data never leaves the private backbone. The call, the transcript, the embeddings, and the search all run on Telnyx infrastructure.

Example use cases

• Voice AI agents that need to recall details from previous calls to maintain continuity across interactions.
• Compliance and audit teams that need to search call transcripts for specific topics, phrases, or regulatory keywords.
• Customer support operations building RAG pipelines from call history, giving agents context from past conversations without manual lookup.
• Healthcare and financial services teams that need indexed, searchable call records without routing sensitive transcripts through third-party vector databases.

Getting started

1. In Mission Control, navigate to Voice → SIP Connections and select your connection.
2. Enable Conversation Persistence on the connection settings.
3. After calls complete, search your indexed history using GET /v2/ai/conversation_histories with a semantic query and optional filters.
4. Use the chunked, relevance-scored results as RAG context in your agent logic.

Learn more in the Conversation History documentation or the Telnyx Inference docs.