Skip to content

Prototype for reasoning tokens #59

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
trask merged 2 commits into from
Apr 8, 2026
Merged

Prototype for reasoning tokens #59

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e0b48e2
Prototype for reasoning tokens
trask Apr 5, 2026
aa615da
Merge branch 'main' into prototype-reasoning-tokens
trask Apr 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions genai_otel_conformance/specs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,6 +64,7 @@ def attrs_for_requirement_level(self, level: RequirementLevel) -> tuple[str, ...
"gen_ai.usage.cache_read.input_tokens",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
]
_INFERENCE_OPT_IN = [
"gen_ai.input.messages",
Expand Down
7 changes: 7 additions & 0 deletions tests/dotnet/azure-openai/Program.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -167,6 +167,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", completion.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", completion.Usage.OutputTokenCount);
if (completion.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", completion.Usage.OutputTokenDetails.ReasoningTokenCount);
}

var content = completion.Content[0].Text;
Expand All @@ -191,6 +193,7 @@ static async Task RunPrototype()
["gen_ai.response.finish_reasons"] = new[] { completion.FinishReason.ToString() },
["gen_ai.usage.input_tokens"] = completion.Usage?.InputTokenCount,
["gen_ai.usage.output_tokens"] = completion.Usage?.OutputTokenCount,
["gen_ai.usage.reasoning.output_tokens"] = completion.Usage?.OutputTokenDetails?.ReasoningTokenCount > 0 ? completion.Usage.OutputTokenDetails.ReasoningTokenCount : null,
["gen_ai.input.messages"] = inputMessagesJson,
["gen_ai.output.messages"] = outputMessagesJson,
["server.address"] = endpoint.Host,
Expand Down Expand Up @@ -228,6 +231,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", update.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", update.Usage.OutputTokenCount);
if (update.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", update.Usage.OutputTokenDetails.ReasoningTokenCount);
}
}

Expand Down Expand Up @@ -301,6 +306,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", completion.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", completion.Usage.OutputTokenCount);
if (completion.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", completion.Usage.OutputTokenDetails.ReasoningTokenCount);
}

if (completion.ToolCalls.Count > 0)
Expand Down
1 change: 1 addition & 0 deletions tests/dotnet/azure-openai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
"server.address",
"server.port"
]
Expand Down
7 changes: 7 additions & 0 deletions tests/dotnet/extensions-ai/Program.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", completion.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", completion.Usage.OutputTokenCount);
if (completion.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", completion.Usage.OutputTokenDetails.ReasoningTokenCount);
}

var content = completion.Content[0].Text;
Expand All @@ -190,6 +192,7 @@ static async Task RunPrototype()
["gen_ai.response.finish_reasons"] = new[] { completion.FinishReason.ToString() },
["gen_ai.usage.input_tokens"] = completion.Usage?.InputTokenCount,
["gen_ai.usage.output_tokens"] = completion.Usage?.OutputTokenCount,
["gen_ai.usage.reasoning.output_tokens"] = completion.Usage?.OutputTokenDetails?.ReasoningTokenCount > 0 ? completion.Usage.OutputTokenDetails.ReasoningTokenCount : null,
["gen_ai.input.messages"] = inputMessagesJson,
["gen_ai.output.messages"] = outputMessagesJson,
["server.address"] = endpoint.Host,
Expand Down Expand Up @@ -227,6 +230,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", update.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", update.Usage.OutputTokenCount);
if (update.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", update.Usage.OutputTokenDetails.ReasoningTokenCount);
}
}

Expand Down Expand Up @@ -300,6 +305,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", completion.Usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", completion.Usage.OutputTokenCount);
if (completion.Usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", completion.Usage.OutputTokenDetails.ReasoningTokenCount);
}

if (completion.ToolCalls.Count > 0)
Expand Down
1 change: 1 addition & 0 deletions tests/dotnet/extensions-ai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
"server.address",
"server.port"
]
Expand Down
11 changes: 11 additions & 0 deletions tests/dotnet/semantic-kernel/Program.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -193,6 +193,7 @@ static async Task RunPrototype()
string finishReason = null;
int? inputTokens = null;
int? outputTokens = null;
int? reasoningTokens = null;
if (metadata != null)
{
if (metadata.TryGetValue("Id", out var id) && id != null)
Expand All @@ -213,6 +214,11 @@ static async Task RunPrototype()
outputTokens = usage.OutputTokenCount;
activity?.SetTag("gen_ai.usage.input_tokens", inputTokens);
activity?.SetTag("gen_ai.usage.output_tokens", outputTokens);
if (usage.OutputTokenDetails?.ReasoningTokenCount > 0)
{
reasoningTokens = usage.OutputTokenDetails.ReasoningTokenCount;
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", reasoningTokens);
}
}
}

Expand All @@ -236,6 +242,7 @@ static async Task RunPrototype()
["gen_ai.response.finish_reasons"] = finishReason != null ? new[] { finishReason } : null,
["gen_ai.usage.input_tokens"] = inputTokens,
["gen_ai.usage.output_tokens"] = outputTokens,
["gen_ai.usage.reasoning.output_tokens"] = reasoningTokens,
["gen_ai.input.messages"] = inputMessagesJson,
["gen_ai.output.messages"] = outputMessagesJson,
["server.address"] = endpoint.Host,
Expand Down Expand Up @@ -279,6 +286,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", usage.OutputTokenCount);
if (usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", usage.OutputTokenDetails.ReasoningTokenCount);
}
}
}
Expand Down Expand Up @@ -347,6 +356,8 @@ static async Task RunPrototype()
{
activity?.SetTag("gen_ai.usage.input_tokens", usage.InputTokenCount);
activity?.SetTag("gen_ai.usage.output_tokens", usage.OutputTokenCount);
if (usage.OutputTokenDetails?.ReasoningTokenCount > 0)
activity?.SetTag("gen_ai.usage.reasoning.output_tokens", usage.OutputTokenDetails.ReasoningTokenCount);
}
}

Expand Down
1 change: 1 addition & 0 deletions tests/dotnet/semantic-kernel/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
"server.address",
"server.port"
],
Expand Down
3 changes: 2 additions & 1 deletion tests/java/openai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,8 @@
"gen_ai.response.model",
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens"
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens"
],
"embeddings": [
"gen_ai.operation.name",
Expand Down
16 changes: 16 additions & 0 deletions tests/java/openai/src/main/java/com/example/openaitest/OpenAiPrototypeTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -89,6 +89,10 @@ static void runChat(OpenAIClient client) {
completion.usage().ifPresent(usage -> {
span.setAttribute(longKey("gen_ai.usage.input_tokens"), usage.promptTokens());
span.setAttribute(longKey("gen_ai.usage.output_tokens"), usage.completionTokens());
usage.completionTokensDetails().ifPresent(details -> {
details.reasoningTokens().filter(r -> r > 0).ifPresent(reasoning ->
span.setAttribute(longKey("gen_ai.usage.reasoning.output_tokens"), reasoning));
});
});

String content = choice.message().content().orElse("");
Expand Down Expand Up @@ -132,6 +136,10 @@ static void runChat(OpenAIClient client) {
logBuilder
.setAttribute(longKey("gen_ai.usage.input_tokens"), usage.promptTokens())
.setAttribute(longKey("gen_ai.usage.output_tokens"), usage.completionTokens());
usage.completionTokensDetails().ifPresent(details -> {
details.reasoningTokens().filter(r -> r > 0).ifPresent(reasoning ->
logBuilder.setAttribute(longKey("gen_ai.usage.reasoning.output_tokens"), reasoning));
});
});
logBuilder.emit();

Expand Down Expand Up @@ -173,6 +181,10 @@ static void runChatStreaming(OpenAIClient client) {
chunk.usage().ifPresent(usage -> {
span.setAttribute(longKey("gen_ai.usage.input_tokens"), usage.promptTokens());
span.setAttribute(longKey("gen_ai.usage.output_tokens"), usage.completionTokens());
usage.completionTokensDetails().ifPresent(details -> {
details.reasoningTokens().filter(r -> r > 0).ifPresent(reasoning ->
span.setAttribute(longKey("gen_ai.usage.reasoning.output_tokens"), reasoning));
});
});
});
} catch (Exception e) {
Expand Down Expand Up @@ -238,6 +250,10 @@ static void runChatToolCall(OpenAIClient client) {
completion.usage().ifPresent(usage -> {
span.setAttribute(longKey("gen_ai.usage.input_tokens"), usage.promptTokens());
span.setAttribute(longKey("gen_ai.usage.output_tokens"), usage.completionTokens());
usage.completionTokensDetails().ifPresent(details -> {
details.reasoningTokens().filter(r -> r > 0).ifPresent(reasoning ->
span.setAttribute(longKey("gen_ai.usage.reasoning.output_tokens"), reasoning));
});
});

List<ChatCompletionMessageToolCall> toolCalls =
Expand Down
1 change: 1 addition & 0 deletions tests/js/azure-openai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
"server.address",
"server.port"
],
Expand Down
12 changes: 12 additions & 0 deletions tests/js/azure-openai/test_prototype.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,9 @@ async function main() {
if (resp.usage) {
span.setAttribute("gen_ai.usage.input_tokens", resp.usage.prompt_tokens);
span.setAttribute("gen_ai.usage.output_tokens", resp.usage.completion_tokens);
if (resp.usage.completion_tokens_details?.reasoning_tokens) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", resp.usage.completion_tokens_details.reasoning_tokens);
}
}

// Emit inference operation details event
Expand All @@ -62,6 +65,7 @@ async function main() {
"gen_ai.response.finish_reasons": resp.choices.map(c => c.finish_reason),
"gen_ai.usage.input_tokens": resp.usage?.prompt_tokens,
"gen_ai.usage.output_tokens": resp.usage?.completion_tokens,
"gen_ai.usage.reasoning.output_tokens": resp.usage?.completion_tokens_details?.reasoning_tokens || undefined,
"gen_ai.input.messages": JSON.stringify(
messages.map(m => ({ role: m.role, parts: [{ type: "text", content: m.content }] }))
),
Expand Down Expand Up @@ -105,6 +109,7 @@ async function main() {
let finishReason = "";
let inputTokens = 0;
let outputTokens = 0;
let reasoningTokens: number | undefined;
for await (const chunk of stream) {
if (chunk.choices[0]?.delta?.content) {
text += chunk.choices[0].delta.content;
Expand All @@ -117,6 +122,9 @@ async function main() {
if (chunk.usage) {
inputTokens = chunk.usage.prompt_tokens;
outputTokens = chunk.usage.completion_tokens;
if (chunk.usage.completion_tokens_details?.reasoning_tokens) {
reasoningTokens = chunk.usage.completion_tokens_details.reasoning_tokens;
}
}
}
span.setAttribute("gen_ai.response.model", model);
Expand All @@ -126,6 +134,7 @@ async function main() {
}
if (inputTokens) span.setAttribute("gen_ai.usage.input_tokens", inputTokens);
if (outputTokens) span.setAttribute("gen_ai.usage.output_tokens", outputTokens);
if (reasoningTokens) span.setAttribute("gen_ai.usage.reasoning.output_tokens", reasoningTokens);
console.log(` -> ${text.slice(0, 60)}`);
span.end();
});
Expand Down Expand Up @@ -168,6 +177,9 @@ async function main() {
if (resp.usage) {
span.setAttribute("gen_ai.usage.input_tokens", resp.usage.prompt_tokens);
span.setAttribute("gen_ai.usage.output_tokens", resp.usage.completion_tokens);
if (resp.usage.completion_tokens_details?.reasoning_tokens) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", resp.usage.completion_tokens_details.reasoning_tokens);
}
}
const toolCall = resp.choices[0].message.tool_calls?.[0];
if (toolCall?.type === "function") {
Expand Down
1 change: 1 addition & 0 deletions tests/js/openai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens",
"server.address",
"server.port"
],
Expand Down
12 changes: 12 additions & 0 deletions tests/js/openai/test_prototype.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,9 @@ async function main() {
if (resp.usage) {
span.setAttribute("gen_ai.usage.input_tokens", resp.usage.prompt_tokens);
span.setAttribute("gen_ai.usage.output_tokens", resp.usage.completion_tokens);
if (resp.usage.completion_tokens_details?.reasoning_tokens) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", resp.usage.completion_tokens_details.reasoning_tokens);
}
}

// Emit inference operation details event
Expand All @@ -58,6 +61,7 @@ async function main() {
"gen_ai.response.finish_reasons": resp.choices.map(c => c.finish_reason),
"gen_ai.usage.input_tokens": resp.usage?.prompt_tokens,
"gen_ai.usage.output_tokens": resp.usage?.completion_tokens,
"gen_ai.usage.reasoning.output_tokens": resp.usage?.completion_tokens_details?.reasoning_tokens || undefined,
"gen_ai.input.messages": JSON.stringify(
messages.map(m => ({ role: m.role, parts: [{ type: "text", content: m.content }] }))
),
Expand Down Expand Up @@ -101,6 +105,7 @@ async function main() {
let finishReason = "";
let inputTokens = 0;
let outputTokens = 0;
let reasoningTokens: number | undefined;
for await (const chunk of stream) {
if (chunk.choices[0]?.delta?.content) {
text += chunk.choices[0].delta.content;
Expand All @@ -113,6 +118,9 @@ async function main() {
if (chunk.usage) {
inputTokens = chunk.usage.prompt_tokens;
outputTokens = chunk.usage.completion_tokens;
if (chunk.usage.completion_tokens_details?.reasoning_tokens) {
reasoningTokens = chunk.usage.completion_tokens_details.reasoning_tokens;
}
}
}
span.setAttribute("gen_ai.response.model", model);
Expand All @@ -122,6 +130,7 @@ async function main() {
}
if (inputTokens) span.setAttribute("gen_ai.usage.input_tokens", inputTokens);
if (outputTokens) span.setAttribute("gen_ai.usage.output_tokens", outputTokens);
if (reasoningTokens) span.setAttribute("gen_ai.usage.reasoning.output_tokens", reasoningTokens);
console.log(` -> ${text.slice(0, 60)}`);
span.end();
});
Expand Down Expand Up @@ -164,6 +173,9 @@ async function main() {
if (resp.usage) {
span.setAttribute("gen_ai.usage.input_tokens", resp.usage.prompt_tokens);
span.setAttribute("gen_ai.usage.output_tokens", resp.usage.completion_tokens);
if (resp.usage.completion_tokens_details?.reasoning_tokens) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", resp.usage.completion_tokens_details.reasoning_tokens);
}
}
const toolCall = resp.choices[0].message.tool_calls?.[0];
if (toolCall?.type === "function") {
Expand Down
3 changes: 2 additions & 1 deletion tests/js/vertexai/data-prototype.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,8 @@
"gen_ai.response.model",
"gen_ai.tool.definitions",
"gen_ai.usage.input_tokens",
"gen_ai.usage.output_tokens"
"gen_ai.usage.output_tokens",
"gen_ai.usage.reasoning.output_tokens"
]
},
"events": {
Expand Down
7 changes: 7 additions & 0 deletions tests/js/vertexai/test_prototype.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,9 @@ async function main() {
if (resp.usageMetadata.candidatesTokenCount) {
span.setAttribute("gen_ai.usage.output_tokens", resp.usageMetadata.candidatesTokenCount);
}
if ((resp.usageMetadata as any).thoughtsTokenCount) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", (resp.usageMetadata as any).thoughtsTokenCount);
}
}
const text = candidate?.content?.parts?.[0]?.text ?? "";

Expand All @@ -97,6 +100,7 @@ async function main() {
"gen_ai.response.finish_reasons": candidate?.finishReason ? [candidate.finishReason] : undefined,
"gen_ai.usage.input_tokens": resp.usageMetadata?.promptTokenCount,
"gen_ai.usage.output_tokens": resp.usageMetadata?.candidatesTokenCount,
"gen_ai.usage.reasoning.output_tokens": (resp.usageMetadata as any)?.thoughtsTokenCount || undefined,
"gen_ai.input.messages": JSON.stringify([
{ role: "user", parts: [{ type: "text", content: userMessage }] }
]),
Expand Down Expand Up @@ -187,6 +191,9 @@ async function main() {
if (aggregated.usageMetadata.candidatesTokenCount) {
span.setAttribute("gen_ai.usage.output_tokens", aggregated.usageMetadata.candidatesTokenCount);
}
if ((aggregated.usageMetadata as any).thoughtsTokenCount) {
span.setAttribute("gen_ai.usage.reasoning.output_tokens", (aggregated.usageMetadata as any).thoughtsTokenCount);
}
}
console.log(` -> ${text.slice(0, 60)}`);
span.end();
Expand Down
Loading