Skip to content

remove tokenizer deps #5026

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

Open
chitalian wants to merge 5 commits into
base: main
Choose a base branch
from
Open

remove tokenizer deps #5026

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
ae8c93b
remove tokenizer deps
chitalian Oct 8, 2025
9384166
Merge branch 'main' into justin/remove-tokenizer-deps
chitalian Oct 8, 2025
52b60af
fix bifrost build
chitalian Oct 8, 2025
0933d6c
Merge main into justin/remove-tokenizer-deps
chitalian Nov 20, 2025
ff63d16
fix: Address PR review feedback for tokenizer removal
chitalian Nov 21, 2025
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
3 changes: 0 additions & 3 deletions valhalla/jawn/package.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,6 @@
"main": "index.js",
"license": "MIT",
"dependencies": {
"@anthropic-ai/tokenizer": "^0.0.4",
"@aws-sdk/client-s3": "^3.678.0",
"@aws-sdk/client-sqs": "^3.799.0",
"@aws-sdk/s3-request-presigner": "^3.537.0",
Expand Down Expand Up @@ -52,7 +51,6 @@
"generate-api-key": "^1.0.2",
"gpt-tokenizer": "^2.1.2",
"ioredis": "^5.3.2",
"js-tiktoken": "^1.0.8",
"kafkajs": "^2.2.4",
"lastmile": "^0.4.0",
"llama-api-client": "^0.1.2",
Expand All @@ -71,7 +69,6 @@
"sorted-btree": "^1.8.1",
"stripe": "17.7.0",
"swagger-ui-express": "^5.0.0",
"tiktoken": "^1.0.15",
"typescript": "^5.2.2",
"uuid": "^9.0.1",
"uuidv4": "^6.2.13",
Expand Down
31 changes: 0 additions & 31 deletions valhalla/jawn/src/lib/routers/tokenRouter.ts
View file
Open in desktop

This file was deleted.

74 changes: 24 additions & 50 deletions valhalla/jawn/src/lib/shared/bodyProcessors/anthropicBodyProcessor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,67 +1,41 @@
import {
calculateModel,
getModelFromResponse,
} from "../../../utils/modelMapper";
import { getTokenCountAnthropic } from "../../tokens/tokenCounter";
import { PromiseGenericResult, ok } from "../../../packages/common/result";
import { IBodyProcessor, ParseInput, ParseOutput } from "./IBodyProcessor";

export class AnthropicBodyProcessor implements IBodyProcessor {
public async parse(
parseInput: ParseInput
): PromiseGenericResult<ParseOutput> {
const { responseBody, requestModel, modelOverride } = parseInput;
const { responseBody } = parseInput;
const parsedResponseBody = JSON.parse(responseBody);
const responseModel = getModelFromResponse(parsedResponseBody);
const model = calculateModel(requestModel, responseModel, modelOverride);

if (
model?.includes("claude-3") ||
model?.includes("claude-sonnet-4") ||
model?.includes("claude-opus-4") ||
// for AI SDK
model?.includes("claude-4")
!parsedResponseBody?.usage?.output_tokens ||
!parsedResponseBody?.usage?.input_tokens
) {
if (
!parsedResponseBody?.usage?.output_tokens ||
!parsedResponseBody?.usage?.input_tokens
) {
return ok({
processedBody: parsedResponseBody,
});
} else {
return ok({
processedBody: parsedResponseBody,
usage: {
totalTokens:
parsedResponseBody?.usage?.input_tokens +
parsedResponseBody?.usage?.output_tokens +
parsedResponseBody?.usage?.cache_creation_input_tokens +
parsedResponseBody?.usage?.cache_read_input_tokens,
promptTokens: parsedResponseBody?.usage?.input_tokens,
promptCacheWriteTokens:
parsedResponseBody?.usage?.cache_creation_input_tokens,
promptCacheReadTokens:
parsedResponseBody?.usage?.cache_read_input_tokens,
completionTokens: parsedResponseBody?.usage?.output_tokens,
promptCacheWrite5m:
parsedResponseBody?.usage?.cache_creation?.ephemeral_5m_input_tokens,
promptCacheWrite1h:
parsedResponseBody?.usage?.cache_creation?.ephemeral_1h_input_tokens,
heliconeCalculated: true,
},
});
}
return ok({
processedBody: parsedResponseBody,
});
} else {
const prompt = parsedResponseBody?.prompt ?? "";
const completion = parsedResponseBody?.completion ?? "";
const completionTokens = await getTokenCountAnthropic(completion);
const promptTokens = await getTokenCountAnthropic(prompt);
return ok({
processedBody: parsedResponseBody,
usage: {
totalTokens: promptTokens + completionTokens,
promptTokens: promptTokens,
completionTokens: completionTokens,
totalTokens:
parsedResponseBody?.usage?.input_tokens +
parsedResponseBody?.usage?.output_tokens +
parsedResponseBody?.usage?.cache_creation_input_tokens +
parsedResponseBody?.usage?.cache_read_input_tokens,
Copy link
Contributor

@greptile-apps greptile-apps bot Oct 8, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

style: Consider using nullish coalescing (?? 0) for cache token values to handle undefined values safely in the addition

Prompt To Fix With AI 
This is a comment left during a code review.
Path: valhalla/jawn/src/lib/shared/bodyProcessors/anthropicBodyProcessor.ts
Line: 22:26

Comment:
**style:** Consider using nullish coalescing (`?? 0`) for cache token values to handle undefined values safely in the addition

How can I resolve this? If you propose a fix, please make it concise.

promptTokens: parsedResponseBody?.usage?.input_tokens,
promptCacheWriteTokens:
parsedResponseBody?.usage?.cache_creation_input_tokens,
promptCacheReadTokens:
parsedResponseBody?.usage?.cache_read_input_tokens,
completionTokens: parsedResponseBody?.usage?.output_tokens,
promptCacheWrite5m:
parsedResponseBody?.usage?.cache_creation
?.ephemeral_5m_input_tokens,
promptCacheWrite1h:
parsedResponseBody?.usage?.cache_creation
?.ephemeral_1h_input_tokens,
heliconeCalculated: true,
},
});
Expand Down
92 changes: 28 additions & 64 deletions valhalla/jawn/src/lib/shared/bodyProcessors/anthropicStreamBodyProcessor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,3 @@
import { calculateModel } from "../../../utils/modelMapper";
import { consolidateTextFields } from "../../../utils/streamParser";
import { getTokenCountAnthropic } from "../../tokens/tokenCounter";
import { PromiseGenericResult, ok } from "../../../packages/common/result";
import { IBodyProcessor, ParseInput, ParseOutput } from "./IBodyProcessor";
import { isParseInputJson } from "./helpers";
Expand All @@ -25,9 +22,7 @@ export class AnthropicStreamBodyProcessor implements IBodyProcessor {
});
}

const { responseBody, requestBody, requestModel, modelOverride } =
parseInput;
const model = calculateModel(requestModel, undefined, modelOverride);
const { responseBody } = parseInput;

// Store the original response body for later use
const originalResponseBody = responseBody;
Expand Down Expand Up @@ -110,69 +105,38 @@ export class AnthropicStreamBodyProcessor implements IBodyProcessor {
}

try {
const processedBody = {
...processConsolidatedJsonForClaude3(processedLines),
// Store the original response body
streamed_data: originalResponseBody,
};

if (
model?.includes("claude-3") ||
model?.includes("claude-sonnet-4") ||
model?.includes("claude-opus-4") ||
// for AI SDK
model?.includes("claude-4")
!processedBody?.usage?.output_tokens ||
!processedBody?.usage?.input_tokens
) {
const processedBody = {
...processConsolidatedJsonForClaude3(processedLines),
// Store the original response body
streamed_data: originalResponseBody,
};

if (
!processedBody?.usage?.output_tokens ||
!processedBody?.usage?.input_tokens
) {
return ok({
processedBody: processedBody,
});
} else {
return ok({
processedBody: processedBody,
usage: {
totalTokens:
processedBody?.usage?.input_tokens +
processedBody?.usage?.output_tokens +
processedBody?.usage?.cache_creation_input_tokens +
processedBody?.usage?.cache_read_input_tokens,
promptTokens: processedBody?.usage?.input_tokens,
promptCacheWriteTokens:
processedBody?.usage?.cache_creation_input_tokens,
promptCacheReadTokens:
processedBody?.usage?.cache_read_input_tokens,
completionTokens: processedBody?.usage?.output_tokens,
promptCacheWrite5m:
processedBody?.usage?.cache_creation?.ephemeral_5m_input_tokens,
promptCacheWrite1h:
processedBody?.usage?.cache_creation?.ephemeral_1h_input_tokens,
heliconeCalculated: true,
},
});
}
return ok({
processedBody: processedBody,
});
} else {
const claudeData = {
...processedLines[processedLines.length - 1],
completion: processedLines.map((d) => d.completion).join(""),
};
const completionTokens = await getTokenCountAnthropic(
claudeData.completion
);
const promptTokens = await getTokenCountAnthropic(
JSON.parse(requestBody ?? "{}")?.prompt ?? ""
);
return ok({
processedBody: {
...consolidateTextFields(processedLines),
streamed_data: originalResponseBody,
},
processedBody: processedBody,
usage: {
totalTokens: completionTokens + promptTokens,
promptTokens: promptTokens,
completionTokens: completionTokens,
totalTokens:
processedBody?.usage?.input_tokens +
processedBody?.usage?.output_tokens +
processedBody?.usage?.cache_creation_input_tokens +
processedBody?.usage?.cache_read_input_tokens,
promptTokens: processedBody?.usage?.input_tokens,
promptCacheWriteTokens:
processedBody?.usage?.cache_creation_input_tokens,
promptCacheReadTokens:
processedBody?.usage?.cache_read_input_tokens,
completionTokens: processedBody?.usage?.output_tokens,
promptCacheWrite5m:
processedBody?.usage?.cache_creation?.ephemeral_5m_input_tokens,
promptCacheWrite1h:
processedBody?.usage?.cache_creation?.ephemeral_1h_input_tokens,
heliconeCalculated: true,
},
});
Expand Down
20 changes: 1 addition & 19 deletions valhalla/jawn/src/lib/shared/bodyProcessors/llamaStreamBodyProcessor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,3 @@
import { calculateModel } from "../../../utils/modelMapper";
import { getTokenCountAnthropic } from "../../tokens/tokenCounter";
import { PromiseGenericResult, ok } from "../../../packages/common/result";
import { IBodyProcessor, ParseInput, ParseOutput } from "./IBodyProcessor";
import { isParseInputJson } from "./helpers";
Expand All @@ -14,9 +12,7 @@ export class LlamaStreamBodyProcessor implements IBodyProcessor {
});
}

const { responseBody, requestBody, requestModel, modelOverride } =
parseInput;
const model = calculateModel(requestModel, undefined, modelOverride);
const { responseBody } = parseInput;

const originalResponseBody = responseBody;
const eventLines = responseBody.split("\n");
Expand Down Expand Up @@ -157,20 +153,6 @@ export class LlamaStreamBodyProcessor implements IBodyProcessor {
}
}

if (!usage) {
const completionTokens = await getTokenCountAnthropic(completionText);
const promptTokens = await getTokenCountAnthropic(
JSON.parse(requestBody ?? "{}")?.messages?.map((m: any) => m.content).join("") ?? ""
);

usage = {
totalTokens: completionTokens + promptTokens,
promptTokens: promptTokens,
completionTokens: completionTokens,
heliconeCalculated: true,
};
}

return ok({
processedBody: processedBody,
usage: usage,
Expand Down
1 change: 0 additions & 1 deletion valhalla/jawn/src/lib/shared/bodyProcessors/openAIStreamProcessor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,4 @@
import { consolidateTextFields } from "../../../utils/streamParser";
import { getTokenCountGPT3 } from "../../tokens/tokenCounter";
import { PromiseGenericResult, err, ok } from "../../../packages/common/result";
import { IBodyProcessor, ParseInput, ParseOutput } from "./IBodyProcessor";
import { isParseInputJson } from "./helpers";
Expand Down
61 changes: 0 additions & 61 deletions valhalla/jawn/src/lib/shared/bodyProcessors/vercelStreamProcessor.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,5 @@
import { PromiseGenericResult, ok } from "../../../packages/common/result";
import { IBodyProcessor, ParseInput, ParseOutput } from "./IBodyProcessor";
import { getTokenCountGPT3 } from "../../tokens/tokenCounter";
import { calculateModel } from "../../../utils/modelMapper";

export class VercelStreamProcessor implements IBodyProcessor {
Expand Down Expand Up @@ -105,66 +104,6 @@ export class VercelStreamProcessor implements IBodyProcessor {
requestModel ||
"unknown";

// If usage is all zeros or undefined, calculate tokens as fallback
if (
!usage ||
usage.totalTokens === 0 ||
(usage.promptTokens === 0 && usage.completionTokens === 0)
) {
try {
// Calculate completion tokens
const completionTokens = completionText
? await getTokenCountGPT3(completionText, model)
: 0;
let promptTokens = 0;

// Try to parse the request body to get prompt tokens
if (requestBody) {
try {
const parsedRequest = JSON.parse(requestBody);
// Vercel uses 'messages'
if (
parsedRequest.messages &&
Array.isArray(parsedRequest.messages)
) {
const promptText = parsedRequest.messages
.map((msg: any) => {
if (msg.content) {
// Handle string content
if (typeof msg.content === "string") {
return msg.content;
}
// Handle content array (structured content)
if (Array.isArray(msg.content)) {
return msg.content
.filter((c: any) => c.type === "text")
.map((c: any) => c.text || "")
.join("");
}
}
return "";
})
.join(" ");
promptTokens = promptText
? await getTokenCountGPT3(promptText, model)
: 0;
}
} catch (e) {
console.error("Error parsing request for token calculation", e);
}
}

usage = {
promptTokens,
completionTokens,
totalTokens: promptTokens + completionTokens,
heliconeCalculated: true,
};
} catch (e) {
console.error("Error calculating tokens for Vercel", e);
}
}

try {
// Create OpenAI-compatible response format
const processedBody: any = {
Expand Down
4 changes: 0 additions & 4 deletions valhalla/jawn/src/lib/tokens/gptWorker.js
View file
Open in desktop

This file was deleted.

12 changes: 0 additions & 12 deletions valhalla/jawn/src/lib/tokens/gptWorker.ts
View file
Open in desktop

This file was deleted.

Loading