@voltagent/server-hono@2.0.9

Patch Changes

• #1199 b6813e9 Thanks @omeraplak! - fix: point A2A agent cards at the JSON-RPC endpoint

  A2A agent cards now advertise /a2a/{serverId} instead of the internal
  /.well-known/{serverId}/agent-card.json discovery document. When the card is
  served through the Hono or Elysia integrations, VoltAgent also resolves that
  endpoint to an absolute URL based on the incoming request.

• Updated dependencies [b6813e9]:

  • @voltagent/a2a-server@2.0.3
  • @voltagent/server-core@2.1.13

@voltagent/server-elysia@2.0.7

Patch Changes

• #1199 b6813e9 Thanks @omeraplak! - fix: point A2A agent cards at the JSON-RPC endpoint

  A2A agent cards now advertise /a2a/{serverId} instead of the internal
  /.well-known/{serverId}/agent-card.json discovery document. When the card is
  served through the Hono or Elysia integrations, VoltAgent also resolves that
  endpoint to an absolute URL based on the incoming request.

• Updated dependencies [b6813e9]:

  • @voltagent/a2a-server@2.0.3
  • @voltagent/server-core@2.1.13

@voltagent/server-core@2.1.13

Patch Changes

• #1199 b6813e9 Thanks @omeraplak! - fix: point A2A agent cards at the JSON-RPC endpoint

  A2A agent cards now advertise /a2a/{serverId} instead of the internal
  /.well-known/{serverId}/agent-card.json discovery document. When the card is
  served through the Hono or Elysia integrations, VoltAgent also resolves that
  endpoint to an absolute URL based on the incoming request.

@voltagent/a2a-server@2.0.3

Patch Changes

• #1199 b6813e9 Thanks @omeraplak! - fix: point A2A agent cards at the JSON-RPC endpoint

  A2A agent cards now advertise /a2a/{serverId} instead of the internal
  /.well-known/{serverId}/agent-card.json discovery document. When the card is
  served through the Hono or Elysia integrations, VoltAgent also resolves that
  endpoint to an absolute URL based on the incoming request.

@voltagent/serverless-hono@2.0.10

Patch Changes

• #1191 a21275f Thanks @ravyg! - fix(serverless-hono): defer waitUntil cleanup to prevent tool crashes in Cloudflare Workers

  The finally block in toCloudflareWorker(), toVercelEdge(), and toDeno() was calling cleanup() immediately when the Response was returned, before streaming and tool execution completed. This cleared the global ___voltagent_wait_until while tools were still using it, causing crashes with time-consuming tools.

  Cleanup is now deferred through the platform's own waitUntil() so it runs only after all pending background work has settled.

• Updated dependencies [19fa54b]:

  • @voltagent/server-core@2.1.12

@voltagent/server-core@2.1.12

Patch Changes

• #1189 19fa54b Thanks @pandego! - Fix the development console-access bypass for Request-based WebSocket paths using ?dev=true.

• Updated dependencies [0dc2935]:

  • @voltagent/core@2.7.0

@voltagent/core@2.7.0

Minor Changes

• #1192 0dc2935 Thanks @ravyg! - feat(core): add prepareStep to AgentOptions for per-step tool control

  Surfaces the AI SDK's prepareStep callback as a top-level AgentOptions property so users can set a default step preparation callback at agent creation time. Per-call prepareStep in method options overrides the agent-level default.

  This enables controlling tool availability, tool choice, and other step settings on a per-step basis without passing prepareStep on every call.

@voltagent/server-core@2.1.11

Patch Changes

• #1183 b48f107 Thanks @omeraplak! - feat: persist selected assistant message metadata to memory

  You can enable persisted assistant message metadata at the agent level or per request.

  const result = await agent.streamText("Hello", {
    memory: {
      userId: "user-1",
      conversationId: "conv-1",
      options: {
        messageMetadataPersistence: {
          usage: true,
          finishReason: true,
        },
      },
    },
  });

  With this enabled, fetching messages from memory returns assistant UIMessage.metadata
  with fields like usage and finishReason, not just stream-time metadata.

  REST API requests can enable the same behavior with options.memory.options:

  curl -X POST http://localhost:3141/agents/assistant/text \
    -H "Content-Type: application/json" \
    -d '{
      "input": "Hello",
      "options": {
        "memory": {
          "userId": "user-1",
          "conversationId": "conv-1",
          "options": {
            "messageMetadataPersistence": {
              "usage": true,
              "finishReason": true
            }
          }
        }
      }
    }'

• Updated dependencies [b48f107, 195155b]:

  • @voltagent/core@2.6.14

@voltagent/core@2.6.14

Patch Changes

• #1183 b48f107 Thanks @omeraplak! - feat: persist selected assistant message metadata to memory

  You can enable persisted assistant message metadata at the agent level or per request.

  const result = await agent.streamText("Hello", {
    memory: {
      userId: "user-1",
      conversationId: "conv-1",
      options: {
        messageMetadataPersistence: {
          usage: true,
          finishReason: true,
        },
      },
    },
  });

  With this enabled, fetching messages from memory returns assistant UIMessage.metadata
  with fields like usage and finishReason, not just stream-time metadata.

  REST API requests can enable the same behavior with options.memory.options:

  curl -X POST http://localhost:3141/agents/assistant/text \
    -H "Content-Type: application/json" \
    -d '{
      "input": "Hello",
      "options": {
        "memory": {
          "userId": "user-1",
          "conversationId": "conv-1",
          "options": {
            "messageMetadataPersistence": {
              "usage": true,
              "finishReason": true
            }
          }
        }
      }
    }'

• #1167 195155b Thanks @octo-patch! - fix: use OpenAI-compatible adapter for MiniMax provider

@voltagent/core@2.6.13

Patch Changes

• #1172 8cb2aa5 Thanks @omeraplak! - fix: tighten prompt-context usage telemetry
  • redact nested large binary fields when estimating prompt context usage
  • preserve circular-reference detection when serializing nested prompt message content
  • exclude runtime-only tool metadata from tool schema token estimates
  • avoid emitting cached and reasoning token span attributes when their values are zero