What's Changed

• anthropic beta for vertex-ai by @narengogi in #1356
• fix: add tool_use, thinking and redacted_thinking content_block_start event handling in Bedrock messages transformer by @VisargD in #1362
• 1.12.3 by @VisargD in #1365

Full Changelog: v1.12.2...v1.12.3

What's Changed

• fix: tools cache_control parameter mapping for bedrock unified messages request transformer by @VisargD in #1360
• 1.12.2 by @VisargD in #1361

Full Changelog: v1.12.1...v1.12.2

What's Changed

• cleaner implementation for image edits route by @narengogi in #1347
• 1.12.1 by @VisargD in #1351

Full Changelog: v1.12.0...v1.12.1

What's Changed

• fix: update responseService to conditionally delete content-encoding header by @roh26it in #1334
• support empty google response by @narengogi in #1339
• chore: realtime api improvements by @b4s36t4 in #1340
• fix: remove the regex replace plugin with the correct regex function parsing by @siddharthsambharia-portkey in #1329
• fix: allow inference profile application when uploading file by @b4s36t4 in #1323
• bug: include stream_options for openai text completion handler by @ianbbqzy in #1344
• feat: add Allowed Request Types plugin to control request processing by @roh26it in #1335
• provider: add NextBit (OpenAI-compatible /chat/completions, /completi… by @MarcNB256 in #1331
• fix: accept scope from request headers to allow azure serverless infe… by @b4s36t4 in #1330
• support performance config for bedrock and apac cross region inferencing profiles by @narengogi in #1326
• add a new unified route for count_tokens endpoint by @narengogi in #1293
• unified route handler for image edits by @narengogi in #1333
• 1.12.0 by @VisargD in #1346

New Contributors

• @ianbbqzy made their first contribution in #1344
• @MarcNB256 made their first contribution in #1331

Full Changelog: v1.11.3...v1.12.0

What's Changed

• fix: portkey plugin response handling by @VisargD in #1299
• Remove redundant apk update calls in Dockerfile by @PeterDaveHello in #1123
• fix: add provider options for azure auth mode by @b4s36t4 in #1298
• fix anthropic streaming tool index increment by @narengogi in #1284
• make created timestamp as unix timestamp in seconds instead of milliseconds by @narengogi in #1290
• feat: improve support for the DashScope provider by @horochx in #1196
• finish reason mapping part 2 by @narengogi in #1214
• support nano banana in strict open ai compliance false mode by @narengogi in #1307
• feat/regex-replace-guardrail by @siddharthsambharia-portkey in #1312
• support video on bedrock by @narengogi in #1304
• integration: add tripo3d provider by @dd-eg-user in #1296
• Support url in conditional router by @ayush-portkey in #1317
• model whitelist + metadata by @vrushankportkey in #1302
• fix: handle stream close failures by @horochx in #1306
• Feat/add 3d model ais by @VisargD in #1318
• mistral on vertex by @narengogi in #1314
• fix: build errors by @VisargD in #1319
• 1.11.3 by @VisargD in #1320
• fix: handle undefined choice in together ai finish reason transform by @VisargD in #1321

New Contributors

• @PeterDaveHello made their first contribution in #1123
• @dd-eg-user made their first contribution in #1296
• @ayush-portkey made their first contribution in #1317

Full Changelog: v1.11.2...v1.11.3

What's Changed

• add few new parameters for gpt-5 by @narengogi in #1273
• feat: add 302AI provider integration by @shiwo6324 in #1235
• Allow getting and setting content part for /messages routes by @roh26it in #1271
• fix params for openrouter reasoning by @narengogi in #1280
• Allow failures when PII guardrail fails, manage blank agents in requests by @roh26it in #1277
• fix: handle complex nested schemas in gemini function calling & structured outputs by @pnkvalavala in #1233
• Users should be able to query models in Portkey through the /models endpoint by @roh26it in #1282
• chore: remove unused providers handler import by @VisargD in #1287
• 1.11.2 by @VisargD in #1288

New Contributors

• @shiwo6324 made their first contribution in #1235
• @pnkvalavala made their first contribution in #1233

Full Changelog: v1.11.1...v1.11.2

What's Changed

• remove unintended change in request header construction where accept encoding headers are being forwarded by @narengogi in #1262
• fix: remove headers from webhook plugin response object by @VisargD in #1251
• mark qdrant as a valid provider by @narengogi in #1247
• fix: project id assignment logic [vertex] by @arturfromtabnine in #1243
• fix: assumed role flow for bedrock inference profile base model fetch by @VisargD in #1252
• [New Plugin] WalledAi by @indranil786 in #1198
• fix: mistral-ai /chat/completitions missing "usage" in stream response by @MateuszKepczynski in #1260
• fix: transform x-ai error response to open-ai format by @b4s36t4 in #1250
• Fix : WalledAI guardrail parameters by @indranil786 in #1263
• fix max_completion_tokens mapping for azure foundry by @narengogi in #1209
• Integration: Bytez Chat Model Provider by @inf3rnus in #1175
• chore: use provider options instead of mutating request body for bedr… by @VisargD in #1265
• fix: add missing parameter for bedrock and vertex transformer calls by @VisargD in #1266
• 1.11.1 by @VisargD in #1267

New Contributors

• @indranil786 made their first contribution in #1198
• @MateuszKepczynski made their first contribution in #1260
• @inf3rnus made their first contribution in #1175

Full Changelog: v1.11.0...v1.11.1

What's Changed

• chore: remove redundant console logs by @VisargD in #1226
• chore: npm update @portkey-ai/mustache by @VisargD in #1231
• refactored tryPost by @roh26it in #1121
• fix: stream cache response mapping by @VisargD in #1236
• Feature: Beta support for /v1/messages route by @narengogi in #1168
• onStatusCodes is respected during fallbacks by @roh26it in #1239
• fix: handle provider error logs addition by @VisargD in #1241
• add support for response format in deepseek by @narengogi in #1238
• chore: batch & fine-tune improvements by @b4s36t4 in #1089
• fix: ignore camel case conversion for model details by @VisargD in #1244
• chore: remove response text check from hooks middleware for output ho… by @VisargD in #1245
• fix: set transformed request body in logs service requestParams field by @VisargD in #1246
• Migrate from pyautogen to ag2 Library by @AG2AI-Admin in #1230
• 1.11.0 by @VisargD in #1248

New Contributors

• @AG2AI-Admin made their first contribution in #1230

Full Changelog: v1.10.2...v1.11.0

What's Changed

• Fix: OpenAI responses API modalities param mapping by @siddharthsambharia-portkey in #1222
• Fix/cohere embed handle null encoding format by @narengogi in #1213
• [bug] clean up resources during unhandled rejections from inside stream transforms by @narengogi in #1207
• Chore/add-jina-ai-dimension-param by @siddharthsambharia-portkey in #1219
• feat: add krutrim as provider by @zhao-lun in #1201
• fix tokens calculation for bedrock models when cache tokens are present by @narengogi in #1211
• chore: support array content values in portkey plugin checks by @VisargD in #1223
• 1.10.2 by @VisargD in #1225

New Contributors

• @zhao-lun made their first contribution in #1201

Full Changelog: v1.10.1...v1.10.2

What's Changed

• computer use for anthropic on bedrock by @narengogi in #1177
• add support for new websearch options param by @narengogi in #1199
• prompt_tokens support in message [anthropic vertex ai] by @arturfromtabnine in #1187
• feat: multimodal embeddings for bedrock titan and cohere by @narengogi in #1130
• 1.10.1 by @VisargD in #1202

New Contributors

• @arturfromtabnine made their first contribution in #1187

Full Changelog: v1.10.0...v1.10.1