Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

deps: Bump the npm-minor-patch group across 1 directory with 6 updates (#4621)

dependabot[bot]5e4827de12c9
diff --git a/scripts/verify_sdk_provider_peers.mjs b/scripts/verify_sdk_provider_peers.mjs
new file mode 100644
index 000000000..93f476096
--- /dev/null
+++ b/scripts/verify_sdk_provider_peers.mjs
@@ -0,0 +1,96 @@
+import assert from 'node:assert/strict';
+import { createServer } from 'node:http';
+import { createRequire } from 'node:module';
+import { pathToFileURL } from 'node:url';
+import path from 'node:path';
+
+const sdk = path.resolve(process.argv[2]);
+const require = createRequire(path.join(sdk, 'package.json'));
+const { createAnthropic } = require('@ai-sdk/anthropic');
+const { createOpenAI } = require('@ai-sdk/openai');
+const OpenAI = require('openai').default;
+const { withHeadroom } = await import(pathToFileURL(path.join(sdk, 'dist/adapters/vercel-ai.js')));
+const { withHeadroom: withOpenAIHeadroom } = await import(pathToFileURL(path.join(sdk, 'dist/adapters/openai.js')));
+assert.ok(['localhost', '127.0.0.1'].includes(new URL(process.env.HEADROOM_BASE_URL).hostname));
+const calls = [];
+const server = createServer(async (request, response) => {
+ const chunks = [];
+ for await (const chunk of request) chunks.push(chunk);
+ const body = JSON.parse(Buffer.concat(chunks).toString());
+ calls.push({ body, url: request.url, authorization: request.headers.authorization });
+ response.writeHead(200, { 'content-type': 'application/json' });
+ if (request.url.endsWith('/chat/completions')) {
+ response.end(JSON.stringify({
+ id: 'chatcmpl-local', object: 'chat.completion', created: 1, model: 'gpt-4o',
+ choices: [{ index: 0, message: { role: 'assistant', content: 'fixture answer' }, finish_reason: 'stop' }],
+ usage: { prompt_tokens: 12, completion_tokens: 2, total_tokens: 14 },
+ }));
+ return;
+ }
+ response.end(JSON.stringify({
+ id: 'msg-local-structured-proof', type: 'message', role: 'assistant',
+ model: 'claude-3-7-sonnet-20250219',
+ content: [{ type: 'text', text: '{"name":"Ada"}' }],
+ stop_reason: 'end_turn', stop_sequence: null,
+ usage: { input_tokens: 12, output_tokens: 8 },
+ }));
+});
+await new Promise(resolve => server.listen(0, '127.0.0.1', resolve));
+try {
+ if (process.argv[3] !== 'openai-only') {
+ const provider = createAnthropic({
+ apiKey: 'local-fixture-only',
+ baseURL: `http://127.0.0.1:${server.address().port}/v1`,
+ });
+ const model = withHeadroom(provider('claude-3-7-sonnet-20250219'), {
+ baseUrl: process.env.HEADROOM_BASE_URL, fallback: false, retries: 0,
+ model: 'claude-3-7-sonnet-20250219',
+ });
+ const result = await model.doGenerate({
+ prompt: [{ role: 'user', content: [{ type: 'text', text: 'Return the name Ada as JSON.' }] }],
+ maxOutputTokens: 2048,
+ responseFormat: { type: 'json', schema: {
+ type: 'object', properties: { name: { type: 'string' } },
+ required: ['name'], additionalProperties: false,
+ } },
+ providerOptions: { anthropic: { thinking: { type: 'enabled', budgetTokens: 1024 } } },
+ });
+ assert.equal(calls.length, 1);
+ assert.equal(calls[0].body.thinking.type, 'enabled');
+ assert.ok(calls[0].body.tools?.some(tool => tool.name === 'json'));
+ const text = result.content.filter(part => part.type === 'text').map(part => part.text).join('');
+ assert.equal(text, '{"name":"Ada"}');
+ console.log('PASS: real provider + Headroom middleware + real compression HTTP path preserve final JSON text without a tool call');
+ }
+ const beforeOpenAI = calls.length;
+ let credentialCalls = 0;
+ const openai = createOpenAI({
+ apiKey: async () => `local-fixture-${++credentialCalls}`,
+ baseURL: `http://127.0.0.1:${server.address().port}/v1`,
+ });
+ const wrapped = withHeadroom(openai.chat('gpt-4o'), {
+ baseUrl: process.env.HEADROOM_BASE_URL, fallback: false, retries: 0,
+ });
+ for (let index = 1; index <= 2; index++) {
+ const result = await wrapped.doGenerate({
+ prompt: [{ role: 'user', content: [{ type: 'text', text: 'Say hello.' }] }],
+ maxOutputTokens: 128,
+ });
+ assert.equal(result.content.find(part => part.type === 'text').text, 'fixture answer');
+ assert.equal(calls[beforeOpenAI + index - 1].authorization, `Bearer local-fixture-${index}`);
+ }
+ assert.equal(credentialCalls, 2);
+ console.log('PASS: real OpenAI provider resolves async credentials separately for both wrapped requests');
+ const native = withOpenAIHeadroom(new OpenAI({
+ apiKey: 'local-native-fixture', baseURL: `http://127.0.0.1:${server.address().port}/v1`,
+ }), { baseUrl: process.env.HEADROOM_BASE_URL, fallback: false, retries: 0 });
+ const completion = await native.chat.completions.create({
+ model: 'gpt-4o', messages: [{ role: 'user', content: 'Say hello.' }],
+ });
+ assert.equal(completion.choices[0].message.content, 'fixture answer');
+ assert.equal(calls.length, beforeOpenAI + 3);
+ console.log('PASS: native OpenAI SDK chat client completes a real HTTP request through the Headroom adapter');
+} finally {
+ server.closeAllConnections();
+ await new Promise(resolve => server.close(resolve));
+}
diff --git a/scripts/verify_sdk_provider_peers.py b/scripts/verify_sdk_provider_peers.py
new file mode 100644
index 000000000..05164906b
--- /dev/null
+++ b/scripts/verify_sdk_provider_peers.py
@@ -0,0 +1,75 @@
+import os
+import socket
+import subprocess
+import sys
+import tempfile
+import threading
+import time
+from pathlib import Path
+
+os.environ["HEADROOM_SKIP_UPSTREAM_CHECK"] = "1"
+os.environ["HEADROOM_DISTRIBUTION"] = "local"
+os.environ["HF_HUB_DISABLE_TELEMETRY"] = "1"
+root = Path(sys.argv[1]) if len(sys.argv) > 1 else Path(__file__).resolve().parents[1]
+state = tempfile.TemporaryDirectory(prefix="headroom-sdk-proof-")
+os.environ["HEADROOM_WORKSPACE_DIR"] = state.name
+os.environ["HEADROOM_CONFIG_DIR"] = str(Path(state.name) / "config")
+sys.path.insert(0, str(root))
+import uvicorn
+
+from headroom.proxy.server import ProxyConfig, create_app
+
+app = create_app(
+ ProxyConfig(
+ optimize=True,
+ cache_enabled=False,
+ rate_limit_enabled=False,
+ cost_tracking_enabled=False,
+ log_requests=False,
+ ccr_inject_tool=False,
+ ccr_handle_responses=False,
+ ccr_context_tracking=False,
+ image_optimize=False,
+ disable_kompress=True,
+ subscription_tracking_enabled=False,
+ stateless=True,
+ )
+)
+seen = []
+
+
+async def count_compress(request, call_next):
+ response = await call_next(request)
+ if request.url.path == "/v1/compress":
+ seen.append(response.status_code)
+ return response
+
+
+listener = socket.socket()
+listener.bind(("127.0.0.1", 0))
+port = listener.getsockname()[1]
+server = uvicorn.Server(uvicorn.Config(app, log_level="warning"))
+thread = threading.Thread(target=server.run, kwargs={"sockets": [listener]}, daemon=True)
+thread.start()
+try:
+ for _ in range(300):
+ if server.started:
+ break
+ if not thread.is_alive():
+ raise RuntimeError("proxy stopped before startup")
+ time.sleep(0.1)
+ assert server.started
+ env = dict(os.environ, HEADROOM_BASE_URL=f"http://127.0.0.1:{port}")
+ command = ["node", str(Path(__file__).with_suffix(".mjs")), str(root / "sdk/typescript")]
+ if len(sys.argv) > 2:
+ command.append(sys.argv[2])
+ result = subprocess.run(command, env=env)
+ assert seen and all(status == 200 for status in seen), seen
+ print("Real Headroom /v1/compress response:", seen, flush=True)
+ raise SystemExit(result.returncode)
+finally:
+ server.should_exit = True
+ thread.join(timeout=15)
+ listener.close()
+ state.cleanup()
diff --git a/sdk/typescript/package-lock.json b/sdk/typescript/package-lock.json
index 4fb5b71af..09908fe5c 100644
--- a/sdk/typescript/package-lock.json
+++ b/sdk/typescript/package-lock.json
@@ -9,14 +9,14 @@
"version": "0.42.0",
"license": "Apache-2.0",
"devDependencies": {
- "@ai-sdk/anthropic": "^4.0.53",
- "@ai-sdk/openai": "^4.0.60",
- "@ai-sdk/provider": "^4.0.14",
+ "@ai-sdk/anthropic": "^4.0.94",
+ "@ai-sdk/openai": "^4.0.106",
+ "@ai-sdk/provider": "^4.0.35",
"@anthropic-ai/sdk": "^0.133.0",
- "@types/node": "^26.5.0",
- "ai": "^7.0.59",
+ "@types/node": "^26.6.6",
+ "ai": "^7.0.159",
"dotenv": "^18.0.1",
- "openai": "^7.10.0",
+ "openai": "^7.46.0",
"tsup": "^8.0.0",
"typescript": "^6.0.3",
"vitest": "^5.0.0"
@@ -46,14 +46,14 @@
}
},
"node_modules/@ai-sdk/anthropic": {
- "version": "4.0.87",
- "resolved": "https://registry.npmjs.org/@ai-sdk/anthropic/-/anthropic-4.0.87.tgz",
- "integrity": "sha512-x55N7G47mh9p38C2a6qMAzmUxlgKDCXoE9ohRaF4D3hcOGQrYIOZZ/9Iv4Uf9/OxlNDog9Yz4YPX2GL1+otGlg==",
+ "version": "4.0.94",
+ "resolved": "https://registry.npmjs.org/@ai-sdk/anthropic/-/anthropic-4.0.94.tgz",
+ "integrity": "sha512-gD0KmsUZTRuhiYleM2SdMrY0c7m+VeRckLX2VVPnEAlQmaKqSw8+noHiK5ED+SHy4tuDf9VQqcHxlGeObzMNyw==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
- "@ai-sdk/provider": "4.0.33",
- "@ai-sdk/provider-utils": "5.0.67"
+ "@ai-sdk/provider": "4.0.35",
+ "@ai-sdk/provider-utils": "5.0.72"
},
"engines": {
"node": ">=22"
@@ -63,14 +63,14 @@
}
},
"node_modules/@ai-sdk/gateway": {
- "version": "4.0.125",
- "resolved": "https://registry.npmjs.org/@ai-sdk/gateway/-/gateway-4.0.125.tgz",
- "integrity": "sha512-9raNdkENebDxVl4piGntpQgOBHB62K1rC87KceV8kkGOc/g7s6lwbxN7C6QTqXYJaM+h1Lnm64MDgnHAmzUApg==",
+ "version": "4.0.132",
+ "resolved": "https://registry.npmjs.org/@ai-sdk/gateway/-/gateway-4.0.132.tgz",
+ "integrity": "sha512-ppStDEz96g+JEuxGiVTP29OKBjPyYZXRxs38wnq3427U2QbZtxe6TukrgSAAq0n4yaHvSvVAWJY9kBlW8XudYg==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
- "@ai-sdk/provider": "4.0.33",
- "@ai-sdk/provider-utils": "5.0.67",
+ "@ai-sdk/provider": "4.0.35",
+ "@ai-sdk/provider-utils": "5.0.72",
"@vercel/oidc": "3.2.0"
},
"engines": {
@@ -81,14 +81,14 @@
}
},
"node_modules/@ai-sdk/openai": {
- "version": "4.0.101",
- "resolved": "https://registry.npmjs.org/@ai-sdk/openai/-/openai-4.0.101.tgz",
- "integrity": "sha512-GZ57KdfEfdHcM+Kkf1KsXq8T2KM2uG6BwPvWzL25F5uKjozpflol6hjxveKEf9v9YtbaKYi+t1PeaQ1g0jOThw==",
+ "version": "4.0.106",
+ "resolved": "https://registry.npmjs.org/@ai-sdk/openai/-/openai-4.0.106.tgz",
+ "integrity": "sha512-xBhSJCrgeTMZ4kCKABX8BJFtp5ZIJSS2qxvGMALA34TMyR6OVMe93cpU0mGWPFaVwPsWy2DWKb5+x7gkJ9SJ8w==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
- "@ai-sdk/provider": "4.0.33",
- "@ai-sdk/provider-utils": "5.0.67"
+ "@ai-sdk/provider": "4.0.35",
+ "@ai-sdk/provider-utils": "5.0.72"
},
"engines": {
"node": ">=22"
@@ -98,9 +98,9 @@
}
},
"node_modules/@ai-sdk/provider": {
- "version": "4.0.33",
- "resolved": "https://registry.npmjs.org/@ai-sdk/provider/-/provider-4.0.33.tgz",
- "integrity": "sha512-C6ylMuAXjgqKJllp6ODIJwUltK/tm3ZmIj4V2fqENP3Y8gif2DyeJSsTcDwFrXppgEKcrBisi/BWx2tYeoeCAw==",
+ "version": "4.0.35",
+ "resolved": "https://registry.npmjs.org/@ai-sdk/provider/-/provider-4.0.35.tgz",
+ "integrity": "sha512-yBz6fmLS8MSCIzy7NiHEUAq56ZLD4ZA1GR6bpXBxgbwUacP1La7LuwAr/xy9jSzthem4F8Xu2OrhvU/JrE9fYA==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
@@ -111,13 +111,13 @@
}
},
"node_modules/@ai-sdk/provider-utils": {
- "version": "5.0.67",
- "resolved": "https://registry.npmjs.org/@ai-sdk/provider-utils/-/provider-utils-5.0.67.tgz",
- "integrity": "sha512-EUzV6GXutGcesWKOY8tRKgijfuv1LUBpqAquMh5k0n74NCjDbca5d5ZG12qhVtM2X5UwM4B2ZMP2XpnAwZlA8Q==",
+ "version": "5.0.72",
+ "resolved": "https://registry.npmjs.org/@ai-sdk/provider-utils/-/provider-utils-5.0.72.tgz",
+ "integrity": "sha512-M6QAztI84jWumQSvKcGOL91DxhChsWxn3D3gYtV0/7E1TPQg9BmLdIerOolUC/jghQbMlQrHyzp0hqdIhUZ2Rw==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
- "@ai-sdk/provider": "4.0.33",
+ "@ai-sdk/provider": "4.0.35",
"@standard-schema/spec": "^1.1.0",
"@workflow/serde": "4.1.0",
"eventsource-parser": "^3.0.8",
@@ -1340,9 +1340,9 @@
"license": "MIT"
},
"node_modules/@types/node": {
- "version": "26.6.5",
- "resolved": "https://registry.npmjs.org/@types/node/-/node-26.6.5.tgz",
- "integrity": "sha512-uMzsv9stDdofIZBx4UGh4Z9ZgXnGC9seRDscilBcTHAeOFk67MZg/k9bWVfO+bk4qZmG+l52hRnLwUiH2LnJTg==",
+ "version": "26.6.6",
+ "resolved": "https://registry.npmjs.org/@types/node/-/node-26.6.6.tgz",
+ "integrity": "sha512-MzQj+8Kx3XI40OmUAE6Qoh58ZMNstNNMwn2nfd3qWsfe7nDNSqJNzes4njhz6AHwrSHOffSLwXU6yb474Q4aog==",
"dev": true,
"license": "MIT",
"dependencies": {
@@ -1428,15 +1428,15 @@
}
},
"node_modules/ai": {
- "version": "7.0.152",
- "resolved": "https://registry.npmjs.org/ai/-/ai-7.0.152.tgz",
- "integrity": "sha512-cxQaE/17qhgqJcLCespbnyaN3c4IkGuNUtBXKSX1yTvlSsQ7ERV2ygFADMG7d+IWCUuvi7R8DrqAy9DGj0cgkg==",
+ "version": "7.0.159",
+ "resolved": "https://registry.npmjs.org/ai/-/ai-7.0.159.tgz",
+ "integrity": "sha512-4e/BwDoKJlMNYZaqERcTbGte+4TwroN9qKnCprog/XVfD9InzVPlhCviDhN452JIRrRsYg9BT0WsQ2C3Ds2mVQ==",
"dev": true,
"license": "Apache-2.0",
"dependencies": {
- "@ai-sdk/gateway": "4.0.125",
- "@ai-sdk/provider": "4.0.33",
- "@ai-sdk/provider-utils": "5.0.67"
+ "@ai-sdk/gateway": "4.0.132",
+ "@ai-sdk/provider": "4.0.35",
+ "@ai-sdk/provider-utils": "5.0.72"
},
"engines": {
"node": ">=22"
@@ -2150,9 +2150,9 @@
}
},
"node_modules/openai": {
- "version": "7.43.0",
- "resolved": "https://registry.npmjs.org/openai/-/openai-7.43.0.tgz",
- "integrity": "sha512-fBjKaDFG8AVfkGYoIo/E+xNRlNk7nx2nQ9GoQtXfePzPjg4QcT6nejkhXayJnEwSpFGfthOrAPI2KXj5yAqg9A==",
+ "version": "7.46.0",
+ "resolved": "https://registry.npmjs.org/openai/-/openai-7.46.0.tgz",
+ "integrity": "sha512-FmbPcDtc0BTCN7ZQ5SBEnC4lPNMEUjA6w/lRkbh7l9IzWx1Mue6P7OEU2cayp5S5YjLuc7VEUfBCYqoQpb6+1w==",
"dev": true,
"license": "Apache-2.0",
"engines": {
@@ -2162,7 +2162,7 @@
"@aws-sdk/credential-provider-node": ">=3.972.0 <4",
"@smithy/hash-node": ">=4.3.0 <5",
"@smithy/signature-v4": ">=5.4.0 <6",
- "google-auth-library": ">=9.15.1 <11",
+ "google-auth-library": ">=9.15.1 <12",
"undici": ">=5 <9",
"ws": "^8.21.0",
"zod": "^3.25 || ^4.0"
diff --git a/sdk/typescript/package.json b/sdk/typescript/package.json
index 096ead3ed..7099a37be 100644
--- a/sdk/typescript/package.json
+++ b/sdk/typescript/package.json
@@ -120,14 +120,14 @@
}
},
"devDependencies": {
- "@ai-sdk/anthropic": "^4.0.53",
- "@ai-sdk/openai": "^4.0.60",
- "@ai-sdk/provider": "^4.0.14",
+ "@ai-sdk/anthropic": "^4.0.94",
+ "@ai-sdk/openai": "^4.0.106",
+ "@ai-sdk/provider": "^4.0.35",
"@anthropic-ai/sdk": "^0.133.0",
- "@types/node": "^26.5.0",
- "ai": "^7.0.59",
+ "@types/node": "^26.6.6",
+ "ai": "^7.0.159",
"dotenv": "^18.0.1",
- "openai": "^7.10.0",
+ "openai": "^7.46.0",
"tsup": "^8.0.0",
"typescript": "^6.0.3",
"vitest": "^5.0.0"