@ai-sdk/otel
1.0.178 · 20 Nov 2026Apache-2.03.9M / weekRepositoryHomepagenpmAtom feed
169 stable releases, newest first, by release line. Notes are quoted as the maintainers published them.
1.0.178
20 Nov 2026Add telemetry, runtime context and lifecycle callbacks to
experimental_extract
Patch Changes
Add telemetry, runtime context and lifecycle callbacks to
experimental_extract
1.0.177
18 Nov 2026otel: End the span of a model call when a stream retry or a steering interrupt starts a new one
Patch Changes
otel: End the span of a model call when a stream retry or a steering interrupt starts a new one
1.0.175 – 1.0.1762 releases that only update dependenciesai 7.0.176Show each
- 1.0.17617 Nov 2026ai 7.0.176
- 1.0.17517 Nov 2026ai 7.0.175
1.0.174
17 Nov 2026otel: Record
gen_ai.output.messagesparts in the order the model produced themOutput messages previously grouped parts by type (reasoning, then all text joined into one part, then tool calls), so they did not match the same assistant message in the next step's
gen_ai.input.messages. Output parts now follow the content order, text blocks are no longer merged, and file, reasoning file, custom, tool error, and tool approval parts are emitted on both the output and input side. The root span records the content of all steps in order, including text from earlier steps. AddsformatOutputContentand deprecatesformatOutputMessages.otel: Redact Google MCP server credentials in tool definitions
Plain
google.mcp_servertool definitions do not carry a provider-owned
telemetry redaction annotation, for example when they are built by hand or come
from an older@ai-sdk/googleversion. Both OTel integrations now redact everygoogle.mcp_serverheader value in telemetry-only copies, so MCP connector
credentials no longer reachai.prompt.toolsorgen_ai.tool.definitions.
Patch Changes
otel: Record
gen_ai.output.messagesparts in the order the model produced themOutput messages previously grouped parts by type (reasoning, then all text joined into one part, then tool calls), so they did not match the same assistant message in the next step's
gen_ai.input.messages. Output parts now follow the content order, text blocks are no longer merged, and file, reasoning file, custom, tool error, and tool approval parts are emitted on both the output and input side. The root span records the content of all steps in order, including text from earlier steps. AddsformatOutputContentand deprecatesformatOutputMessages.otel: Redact Google MCP server credentials in tool definitions
Plain
google.mcp_servertool definitions do not carry a provider-owned
telemetry redaction annotation, for example when they are built by hand or come
from an older@ai-sdk/googleversion. Both OTel integrations now redact everygoogle.mcp_serverheader value in telemetry-only copies, so MCP connector
credentials no longer reachai.prompt.toolsorgen_ai.tool.definitions.
1.0.169 – 1.0.1735 releases that only update dependenciesai 7.0.173@ai-sdk/provider 4.0.38Show each
- 1.0.17313 Nov 2026ai 7.0.173
- 1.0.17213 Nov 2026ai 7.0.172
- 1.0.17112 Nov 2026ai 7.0.171@ai-sdk/provider 4.0.38
- 1.0.17011 Nov 2026ai 7.0.170
- 1.0.16910 Nov 2026ai 7.0.169
1.0.168
10 Nov 2026Restore credential redaction for unannotated OpenAI MCP, Code Interpreter, and shell tool definitions in both OpenTelemetry integrations. This preserves protection for older providers and plain provider tool definitions without changing application or provider arguments.
Patch Changes
Restore credential redaction for unannotated OpenAI MCP, Code Interpreter, and shell tool definitions in both OpenTelemetry integrations. This preserves protection for older providers and plain provider tool definitions without changing application or provider arguments.
1.0.166
9 Nov 2026Export the GenAI message and system instruction formatting helpers and their output types for reuse outside the telemetry integration.
1.0.164 – 1.0.1652 releases that only update dependenciesai 7.0.165@ai-sdk/provider 4.0.37Show each
- 1.0.1657 Nov 2026ai 7.0.165@ai-sdk/provider 4.0.37
- 1.0.1646 Nov 2026ai 7.0.164
1.0.163
5 Nov 2026Report the actual OpenAI Chat HTTP endpoint (also used by Azure Chat and Bedrock Mantle Chat) through optional request metadata and model-call end events. Record its host and port on OpenTelemetry chat spans without exposing URL credentials, paths, queries, or fragments. Expose only the sanitized HTTP origin in text and object generation results and step/finish callbacks. Providers that do not report an endpoint leave these attributes unset.
Redact Code Interpreter and shell domain secret values, plus MCP authorization and header values, from tool definitions exported by both telemetry integrations, while preserving the credentials in provider requests. Azure tools that share these definitions receive the same redaction.
Patch Changes
Report the actual OpenAI Chat HTTP endpoint (also used by Azure Chat and Bedrock Mantle Chat) through optional request metadata and model-call end events. Record its host and port on OpenTelemetry chat spans without exposing URL credentials, paths, queries, or fragments. Expose only the sanitized HTTP origin in text and object generation results and step/finish callbacks. Providers that do not report an endpoint leave these attributes unset.
Redact Code Interpreter and shell domain secret values, plus MCP authorization and header values, from tool definitions exported by both telemetry integrations, while preserving the credentials in provider requests. Azure tools that share these definitions receive the same redaction.
1.0.162
3 Nov 2026Support a single telemetry operation for durable
WorkflowAgentruns, including the prompt, final output, steps, and tools. Events recorded in the workflow are reported on completion or suspension with their original timestamps. For OpenTelemetry, pass the sameOpenTelemetryIdGeneratorinstance as the tracer provider'sidGeneratorandsamplerand asnew OpenTelemetry({ durable }). This explicit bridge carries parent trace context and sampling across steps, reports model spans live with provider requests nested beneath them, and reconstructs the enclosinginvoke_agenttree when the run completes. Without the bridge, existing live per-model tracing is preserved. Add typed experimental durable context hooks for telemetry integrations; remove reliance on process-global ID-generator detection.
Patch Changes
Support a single telemetry operation for durable
WorkflowAgentruns, including the prompt, final output, steps, and tools. Events recorded in the workflow are reported on completion or suspension with their original timestamps. For OpenTelemetry, pass the sameOpenTelemetryIdGeneratorinstance as the tracer provider'sidGeneratorandsamplerand asnew OpenTelemetry({ durable }). This explicit bridge carries parent trace context and sampling across steps, reports model spans live with provider requests nested beneath them, and reconstructs the enclosinginvoke_agenttree when the run completes. Without the bridge, existing live per-model tracing is preserved. Add typed experimental durable context hooks for telemetry integrations; remove reliance on process-global ID-generator detection.
1.0.161
2 Nov 2026Only updates a dependency:ai 7.0.161
1.0.160
2 Nov 2026Record the
gen_ai.client.inference.durationOpenTelemetry histogram in seconds forgenerateTextandstreamTextmodel calls, including errors, cancellation, and failed stream retry attempts, so configured metric exporters can report inference latency.
Patch Changes
Record the
gen_ai.client.inference.durationOpenTelemetry histogram in seconds forgenerateTextandstreamTextmodel calls, including errors, cancellation, and failed stream retry attempts, so configured metric exporters can report inference latency.
1.0.155 – 1.0.1595 releases that only update dependenciesai 7.0.159@ai-sdk/provider 4.0.35Show each
- 1.0.15930 Oct 2026ai 7.0.159
- 1.0.15830 Oct 2026ai 7.0.158@ai-sdk/provider 4.0.35
- 1.0.15729 Oct 2026ai 7.0.157
- 1.0.15628 Oct 2026ai 7.0.156
- 1.0.15527 Oct 2026ai 7.0.155
1.0.154
26 Oct 2026Add onStepError callback to generateText, streamText, ToolLoopAgent and telemetry integrations, and onError and onAbort to ToolLoopAgent
Patch Changes
Add onStepError callback to generateText, streamText, ToolLoopAgent and telemetry integrations, and onError and onAbort to ToolLoopAgent
1.0.150 – 1.0.1534 releases that only update dependencies@ai-sdk/provider 4.0.34ai 7.0.153Show each
- 1.0.15324 Oct 2026@ai-sdk/provider 4.0.34ai 7.0.153
- 1.0.15223 Oct 2026ai 7.0.152
- 1.0.15123 Oct 2026ai 7.0.151
- 1.0.15022 Oct 2026ai 7.0.150@ai-sdk/provider 4.0.33
1.0.149
22 Oct 2026Add an
agentSpanNameoption toOpenTelemetryto name root agent spans usingfunctionIdor a custom callback. Preserve model-based span names by default.
Patch Changes
Add an
agentSpanNameoption toOpenTelemetryto name root agent spans usingfunctionIdor a custom callback. Preserve model-based span names by default.
1.0.139 – 1.0.14810 releases that only update dependenciesai 7.0.148@ai-sdk/provider 4.0.31Show each
- 1.0.14821 Oct 2026ai 7.0.148
- 1.0.14720 Oct 2026ai 7.0.147
- 1.0.14620 Oct 2026ai 7.0.146@ai-sdk/provider 4.0.31
- 1.0.14519 Oct 2026ai 7.0.145
- 1.0.14419 Oct 2026@ai-sdk/provider 4.0.30ai 7.0.144
- 1.0.14316 Oct 2026ai 7.0.143@ai-sdk/provider 4.0.29
- 1.0.14215 Oct 2026@ai-sdk/provider 4.0.28ai 7.0.142
- 1.0.14115 Oct 2026ai 7.0.141
- 1.0.14014 Oct 2026ai 7.0.140
- 1.0.13913 Oct 2026ai 7.0.139
1.0.138
12 Oct 2026otel: Record decision model response identifiers
Patch Changes
otel: Record decision model response identifiers
1.0.134 – 1.0.1374 releases that only update dependenciesai 7.0.137@ai-sdk/provider 4.0.26Show each
- 1.0.1379 Oct 2026ai 7.0.137
- 1.0.1369 Oct 2026ai 7.0.136
- 1.0.1358 Oct 2026ai 7.0.135@ai-sdk/provider 4.0.26
- 1.0.1348 Oct 2026ai 7.0.134
1.0.133
8 Oct 2026Add ordered text, file, and JSON parts to experimental decision state. Support image input in OpenAI Decisions and language model adapters. Gateway retains its existing string and JSON request format and rejects files, as does TypeSafe AI.
Arrays passed directly as state now contain decision state parts. Wrap JSON arrays in an object or a json part.
Normalize all public state forms into an array of parts before calling decision providers. Providers receive text and JSON objects as text and json parts.
Serialize decision file bytes as base64 in OpenTelemetry state attributes. Model-call spans record normalized state parts, while outer spans retain public state inputs.
Preserve native JSON state in TypeSafe AI when state contains one JSON part. Label shared state in language-model decision prompts.
Patch Changes
Add ordered text, file, and JSON parts to experimental decision state. Support image input in OpenAI Decisions and language model adapters. Gateway retains its existing string and JSON request format and rejects files, as does TypeSafe AI.
Arrays passed directly as state now contain decision state parts. Wrap JSON arrays in an object or a json part.
Normalize all public state forms into an array of parts before calling decision providers. Providers receive text and JSON objects as text and json parts.
Serialize decision file bytes as base64 in OpenTelemetry state attributes. Model-call spans record normalized state parts, while outer spans retain public state inputs.
Preserve native JSON state in TypeSafe AI when state contains one JSON part. Label shared state in language-model decision prompts.
1.0.129 – 1.0.1324 releases that only update dependenciesai 7.0.132@ai-sdk/provider 4.0.24Show each
- 1.0.1327 Oct 2026ai 7.0.132
- 1.0.1317 Oct 2026ai 7.0.131
- 1.0.1307 Oct 2026ai 7.0.130@ai-sdk/provider 4.0.24
- 1.0.1296 Oct 2026ai 7.0.129@ai-sdk/provider 4.0.23
1.0.128
5 Oct 2026Rename
evaluatetodecideotel: Mark span status as
ERRORwhen finish reason iserror
Patch Changes
Rename
evaluatetodecideotel: Mark span status as
ERRORwhen finish reason iserror
A whole project to check? Paste its package.json in the dependency check: every deprecated, outdated or vulnerable dependency, with what to upgrade to. Free, no account.