@juspay/neurolink 12.44.1 → 12.44.3

package/dist/providers/catalog/pareto-inference.json

@@ -16,13 +16,13 @@
             "z-ai/glm-5.3-flash": {
                 "maxOutputTokens": 131072,
                 "pricingPerMTok": {
-                    "input": 0.03,
-                    "output": 0.1,
-                    "cachedInput": 0.006
+                    "input": 0.09,
+                    "output": 0.3,
+                    "cachedInput": 0.018
                 },
                 "vision": false,
                 "status": "production",
-                "description": "Recommended - the only model Pareto serves; GLM 5.3 Flash (Zhipu/Z.ai), run on Pareto's own GPUs. Confirmed live via an unauthenticated GET https://api.paretoinference.com/v1/models on 2026-09-28 (HTTP 200, exactly one id: z-ai/glm-5.3-flash, owned_by z-ai) - no key used, matching the docs' note that GET /models needs no API key. The docs also document a bare alias, glm-5.3-flash, for the same model (https://docs.paretoinference.com/api-reference, retrieved 2026-09-28). Wire is OpenAI Chat Completions: POST /chat/completions with required model + messages ('Use the OpenAI message format'), the answer read from choices[0].message.content, SSE streaming with text in choices[0].delta.content and tool calls in choices[0].delta.tool_calls, a final usage event when stream_options.include_usage is true, and a terminal 'data: [DONE]' (https://docs.paretoinference.com/api-reference, retrieved 2026-09-28). Output tokens (max_tokens / max_completion_tokens - send only one) range 1-131,072, defaulting to 131,072 when both are omitted - that figure is this entry's defaultMaxOutputTokens and maxOutputTokens. Pareto documents no separate context-length figure anywhere: 'We do not disclose the technical details of our GLM 5.3 Flash serving system' (https://docs.paretoinference.com/how-pareto-works, retrieved 2026-09-28), and no context-window number appears on the quickstart, API reference, how-it-works, errors, or pricing pages, so models.defaultContextWindow above (131,072) reuses the documented output ceiling as a conservative floor rather than a vendor-stated context length - correct it once Pareto publishes one or an authenticated probe reveals one; contextWindow is deliberately left unset on this model entry rather than restating that same undocumented floor as a per-model fact. Pricing from https://docs.paretoinference.com/pricing, retrieved 2026-09-28: $0.03 per 1M input tokens, $0.10 per 1M output tokens, $0.006 per 1M cached-input tokens ('a cache hit is not guaranteed'). tools ('Sends OpenAI function tools') and response_format ('Requests a structured response') are both documented request parameters, but the same request-parameters table hedges both of them - plus tool_choice and logprobs - with 'Model support can differ'; there is no model-specific confirmation for z-ai/glm-5.3-flash and no live probe was possible credential-free, so capabilities.tools is 'model-dependent' here and toolsWithStreaming/structuredOutput/structuredOutputWithTools are left false rather than asserted from a hedged parameter table. reasoning_effort ('Sets the reasoning level') is documented as a plain, unhedged request parameter - unlike tools/response_format/tool_choice/logprobs it carries no 'Model support can differ' caveat - which is the basis for capabilities.thinking: true. Vision/image input and an embeddings endpoint are not documented anywhere on docs.paretoinference.com, so vision stays false and models.visionModel is left unset. No live capability probe was run: this is credential-free onboarding, no Pareto API key was ever created, and no POST request was ever sent to the vendor - see evidence.liveMatrix: null."
+                "description": "Recommended - the only model Pareto serves; GLM 5.3 Flash (Zhipu/Z.ai), run on Pareto's own GPUs. Confirmed live via an unauthenticated GET https://api.paretoinference.com/v1/models on 2026-09-28 (HTTP 200, exactly one id: z-ai/glm-5.3-flash, owned_by z-ai) - no key used, matching the docs' note that GET /models needs no API key. The docs also document a bare alias, glm-5.3-flash, for the same model (https://docs.paretoinference.com/api-reference, retrieved 2026-09-28). Wire is OpenAI Chat Completions: POST /chat/completions with required model + messages ('Use the OpenAI message format'), the answer read from choices[0].message.content, SSE streaming with text in choices[0].delta.content and tool calls in choices[0].delta.tool_calls, a final usage event when stream_options.include_usage is true, and a terminal 'data: [DONE]' (https://docs.paretoinference.com/api-reference, retrieved 2026-09-28). Output tokens (max_tokens / max_completion_tokens - send only one) range 1-131,072, defaulting to 131,072 when both are omitted - that figure is this entry's defaultMaxOutputTokens and maxOutputTokens. Pareto documents no separate context-length figure anywhere: 'We do not disclose the technical details of our GLM 5.3 Flash serving system' (https://docs.paretoinference.com/how-pareto-works, retrieved 2026-09-28), and no context-window number appears on the quickstart, API reference, how-it-works, errors, or pricing pages, so models.defaultContextWindow above (131,072) reuses the documented output ceiling as a conservative floor rather than a vendor-stated context length - correct it once Pareto publishes one or an authenticated probe reveals one; contextWindow is deliberately left unset on this model entry rather than restating that same undocumented floor as a per-model fact. Pricing from https://docs.paretoinference.com/pricing, retrieved 2026-10-02: $0.09 per 1M input tokens, $0.30 per 1M output tokens, $0.018 per 1M cached-input tokens ('a cache hit is not guaranteed'); GET https://api.paretoinference.com/v1/models reports the same three figures. The 2026-09-28 onboarding recorded $0.03 / $0.10 / $0.006. The vendor prices this model as a discount to OpenRouter's listing, so it can move; recheck before relying on it. tools ('Sends OpenAI function tools') and response_format ('Requests a structured response') are both documented request parameters, but the same request-parameters table hedges both of them - plus tool_choice and logprobs - with 'Model support can differ'; there is no model-specific confirmation for z-ai/glm-5.3-flash and no live probe was possible credential-free, so capabilities.tools is 'model-dependent' here and toolsWithStreaming/structuredOutput/structuredOutputWithTools are left false rather than asserted from a hedged parameter table. reasoning_effort ('Sets the reasoning level') is documented as a plain, unhedged request parameter - unlike tools/response_format/tool_choice/logprobs it carries no 'Model support can differ' caveat - which is the basis for capabilities.thinking: true. Vision/image input and an embeddings endpoint are not documented anywhere on docs.paretoinference.com, so vision stays false and models.visionModel is left unset. No live capability probe was run: this is credential-free onboarding, no Pareto API key was ever created, and no POST request was ever sent to the vendor - see evidence.liveMatrix: null."
             }
         },
         "topModels": ["z-ai/glm-5.3-flash"]

package/dist/providers/nvidiaNim/client.d.ts

@@ -1,4 +1,4 @@
-import type { AIProviderName } from "../../constants/enums.js";
+import { AIProviderName } from "../../constants/enums.js";
 import type { NeurolinkCredentials, OpenAICompatBuildBodyArgs, OpenAICompatChatRequest } from "../../types/index.js";
 import { OpenAIChatCompletionsProvider } from "../openaiChatCompletionsBase.js";
 declare const isNimFieldRejection: (body: string, field: string) => boolean;

package/dist/providers/nvidiaNim/client.js

@@ -1,4 +1,4 @@
-import { NvidiaNimModels } from "../../constants/enums.js";
+import { AIProviderName, NvidiaNimModels } from "../../constants/enums.js";
 import { AuthenticationError, InvalidModelError, ProviderError, RateLimitError, } from "../../types/index.js";
 import { classifyProviderError, DEFAULT_ERROR_RULES, } from "../../utils/errorClassifier.js";
 import { logger } from "../../utils/logger.js";
@@ -145,8 +145,8 @@ const getDefaultNimModel = () => {
     // pointed out the default lived in four places and that changing only some
     // of them is a no-op — which happened twice while tracking this down. The
     // descriptor is the Factory+Registry convention's source of truth.
-    return getProviderModel("NVIDIA_NIM_MODEL", PROVIDER_DESCRIPTORS_BY_NAME.get("nvidia-nim")
-        ?.defaultModel ?? NvidiaNimModels.GPT_OSS_20B);
+    return getProviderModel("NVIDIA_NIM_MODEL", PROVIDER_DESCRIPTORS_BY_NAME.get(AIProviderName.NVIDIA_NIM)?.defaultModel ??
+        NvidiaNimModels.GPT_OSS_20B);
 };
 /**
  * NVIDIA NIM Provider — native HTTP+SSE, no AI SDK.
@@ -176,7 +176,7 @@ export class NvidiaNimProvider extends OpenAIChatCompletionsProvider {
         const baseURL = credentials?.baseURL ??
             process.env.NVIDIA_NIM_BASE_URL ??
             NVIDIA_NIM_DEFAULT_BASE_URL;
-        super("nvidia-nim", modelName, sdk, { baseURL, apiKey });
+        super(AIProviderName.NVIDIA_NIM, modelName, sdk, { baseURL, apiKey });
         logger.debug("NVIDIA NIM Provider initialized", {
             modelName: this.modelName,
             providerName: this.providerName,
@@ -184,7 +184,7 @@ export class NvidiaNimProvider extends OpenAIChatCompletionsProvider {
         });
     }
     getProviderName() {
-        return "nvidia-nim";
+        return AIProviderName.NVIDIA_NIM;
     }
     getDefaultModel() {
         return getDefaultNimModel();

package/dist/rag/ragIntegration.js

@@ -349,8 +349,9 @@ async function _prepareRAGToolInner(ragConfig, fallbackProvider) {
     // those config fields had no runtime effect and retrieval always used the
     // deterministic hash embedding, which is a lexical fingerprint rather than
     // a semantic space. Index and query must share one embedding space, so the
-    // provider path replaces the hash path wholesale; any provider failure
-    // falls back to the hash for both sides.
+    // provider path replaces the hash path wholesale; a provider failure while
+    // creating the provider or building the index falls back to the hash for
+    // both sides, but one at query time fails that search.
     const wantProviderEmbeddings = Boolean(embeddingProvider || embeddingModel);
     const embedProviderName = embeddingProvider || fallbackProvider || "vertex";
     const embedModelName = embeddingModel || "gemini-2.5-flash";
@@ -450,9 +451,13 @@ async function _prepareRAGToolInner(ragConfig, fallbackProvider) {
                     "rag.top_k": topK ?? 5,
                 },
             }, async (span) => {
-                // Query through the same embedding space the index was built in —
-                // provider embeddings when configured (and healthy), else the hash.
-                const queryEmbedding = await embedFn(query).catch(() => generateSimpleEmbedding(query, EMBEDDING_DIMENSION));
+                // Query through the same embedding space the index was built in.
+                // A failure here must fail the search: swapping in a hash vector
+                // would be scored against provider-space vectors of another
+                // dimension, which cosine reads as 0 for every chunk, so the tool
+                // would return arbitrary chunks as if they were relevant. A hash-built
+                // index never reaches this — its embedFn cannot reject.
+                const queryEmbedding = await embedFn(query);
                 // Fetch more candidates than needed so diversity can select across files and images
                 const loadedSources = fileContents.length + imageChunks.length;
                 const fetchK = loadedSources > 1 ? topK * 3 : topK;

package/docs-site/static/search-index.json

@@ -6564,7 +6564,7 @@
 {"objectID":"1637c68ac1e9cbfce5f16fd46aaa8835701c2f1430482c8fa9910e486a7bb625","title":"Verification status","url":"/docs/getting-started/providers/novita#verification-status","content":"Tier-2 onboarding requires evidence before a provider is accepted, and\npnpm run verify:provider-onboarding gates it in CI. This is what the\ncatalog records for Novita AI:\n\n| Probe                 | Result                                                                                                                                                                                                                                                                                                                                     |\n| --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |\n| Roster                | authenticated GET /openai/v1/models, HTTP 200, 2026-09-05                                                                                                                                                                                                                                                                                  |\n| Auth rejection        | HTTP 401 (FAILED_TO_AUTH), 2026-09-05                                                                                                                                                                                                                                                                                                    |\n| Live capability sweep | 2026-09-05, probed against the default model. Chat, streaming, and a real structured tool call all passed. response_format: json_object succeeded both alone and combined with tools in the same request — confirmed at the SDK level, which is why structuredOutputWithTools is true for this provider, unlike Friendli or Morph. |","hierarchy":{"lvl0":"Getting Started","lvl1":"Novita AI Provider Guide","lvl2":"Verification status","lvl3":""}},
 {"objectID":"a28a46f2a803a788421fbf99c39dcb97d33871e99a7dc45807a98ddce3e15589","title":"Troubleshooting","url":"/docs/getting-started/providers/novita#troubleshooting","content":"| Symptom                                         | Cause                                                                                                | Fix                                                                                             |\n| ----------------------------------------------- | ---------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------- |\n| FAILED_TO_AUTH                                | NOVITA_API_KEY unset, wrong, or not sk_-prefixed                                                 | Rotate the key at https://novita.ai/settings/key-management                                     |\n| MODEL_NOT_FOUND                               | The roster changed since 2026-09-05                                                                  | Pick a current id from https://novita.ai/models or the authenticated /openai/v1/models roster |\n| Truncated or empty content on the default model | zai-org/glm-5.3-flash is a reasoning model and the catalog default max output is only 4,096 tokens | Raise maxTokens for longer or reasoning-heavy completions                                     |","hierarchy":{"lvl0":"Getting Started","lvl1":"Novita AI Provider Guide","lvl2":"Troubleshooting","lvl3":""}},
 {"objectID":"7ce1eb2bda1d99489069643ba7316d9d1e893babf249024cf111547fbe952f66","title":"See also","url":"/docs/getting-started/providers/novita#see-also","content":"Provider setup overview\nAll providers\nTier-2 onboarding\nProvider feature compatibility","hierarchy":{"lvl0":"Getting Started","lvl1":"Novita AI Provider Guide","lvl2":"See also","lvl3":""}},
-{"objectID":"c4d6bc20fbe9d189291d36b241fffab77fac1dcbe5ff66e1df5248840997c55d","title":"NVIDIA NIM Provider Guide","url":"/docs/getting-started/providers/nvidia-nim","content":"NVIDIA NIM Provider Guide\n\nHundreds of optimised AI models on NVIDIA's GPU-accelerated inference platform — or your own self-hosted NIM deployment\n\nOverview\n\nNVIDIA NIM (NVIDIA Inference Microservices) is a managed inference platform that hosts a large catalog of open-weight models — Meta Llama, DeepSeek, Mistral, Microsoft Phi, Google Gemma, and more — all GPU-optimised and served through an OpenAI-compatible API. You can also point NeuroLink at a self-hosted NIM cluster by overriding the base URL.\n\nKey Facts\nHosted base URL: https://integrate.api.nvidia.com/v1\nProtocol: OpenAI-compatible (/v1/chat/completions)\nVision: Yes, on supported models (Llama 3.2 Vision, etc.)\nReasoning: Yes, on Nemotron and DeepSeek-R1 variants\nStreaming: Supported\nTool calling: Supported on most models\nSelf-hosting: Override NVIDIA_NIM_BASE_URL to point at a private NIM cluster\n\nNIM-Specific Extras\n\nNIM supports additional generation parameters beyond the standard OpenAI surface: top_k, min_p, repetition_penalty, min_tokens, and per-model chat_template overrides. The NeuroLink provider automatically passes these via the providerOptions.openai.body mechanism. If a model rejects an unsupported parameter with HTTP 400, the provider retries the request with that parameter stripped.\n\nQuick Start\nGet an API Key\n\nSign up at https://build.nvidia.com and create an API key under API Keys.\nConfigure Environment\n\nAdd to your .env file:\nInstall NeuroLink\nGenerate Your First Response\n\nSupported Models\n\nNIM hosts hundreds of models. NeuroLink ships with these popular models pre-enumerated:\n\nMeta Llama\n\n| Model ID                             | Context | Vision | Reasoning |\n| ------------------------------------ | ------- | ------ | --------- |\n| meta/llama-3.3-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.1-405b-instruct       | 128K    | No     | No        |\n| meta/llama-3.1-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.2-90b-vision-instruct | 128K    | Yes    | No        |\n| meta/llama-3.2-11b-vision-instruct | 128K    | Yes    | No        |\n\nNVIDIA Nemotron (Reasoning)\n\n| Model ID                                 | Context | Vision | Reasoning |\n| ---------------------------------------- | ------- | ------ | --------- |\n| nvidia/llama-3.3-nemotron-super-49b-v1 | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-nano-8b-v1   | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-70b-instruct | 128K    | No     | Yes       |\n\nDeepSeek (Hosted on NIM)\n\n| Model ID                                    | Context | Vision | Reasoning |\n| ------------------------------------------- | ------- | ------ | --------- |\n| deepseek-ai/deepseek-r1                   | 128K    | No     | Yes       |\n| deepseek-ai/deepseek-r1-distill-llama-70b | 128K    | No     | Yes       |\n\nOther Models\n\n| Model ID                                | Context | Notes            |\n| --------------------------------------- | ------- | ---------------- |\n| mistralai/mixtral-8x22b-instruct-v0.1 | 64K     | Large MoE        |\n| mistralai/mixtral-8x7b-instruct-v0.1  | 32K     | Efficient MoE    |\n| microsoft/phi-4                       | 16K     | Compact, capable |\n| google/gemma-3-27b-it                 | 128K    | Google Gemma     |\n\nBrowse the full catalog at https://build.nvidia.com/models. You can pass any model ID via --model or model: — NIM returns 404 for IDs that are not in the catalog.\n\nSDK Usage\n\nBasic Generation\n\nUsing a Specific Model\n\nReasoning with thinkingLevel\n\nReasoning-capable models (Nemotron, DeepSeek-R1) accept a thinking flag via NIM's chat_template_kwargs. Pass thinkingLevel to activate it:\n\nLevels: minimal (no thinking) | low | medium | high\n\nIf the model does not support chat_template_kwargs.thinking, the provider automatically retries without it.\n\nStreaming\n\nPer-Call Credential Override\n\nFor self-hosted NIM clusters, override the base URL per call:\n\nCLI Usage\n\nBasic Commands\n\nProvider Aliases\n\n| Alias        | Example                 |\n| ------------ | ----------------------- |\n| nvidia-nim | --provider nvidia-nim |\n| nvidia     | --provider nvidia     |\n| nim        | --provider nim        |\n\nConfiguration Reference\n\n| Environment Variable            | Required | Default                               | Description                                |\n| ------------------------------- | -------- | ------------------------------------- | ------------------------------------------ |\n| NVIDIA_NIM_API_KEY            | Yes      | —                                     | NVIDIA NIM API key (starts with nvapi-)  |\n| NVIDIA_NIM_MODEL              | No       | meta/llama-3.3-70b-instruct         | Default model                              |\n| NVIDIA_NIM_BASE_URL           | No       | https://integrate.api.nvidia.com/v1 | Base URL (override for self-hosted NIM)    |\n| NVIDIA_NIM_TOP_K              | No       | —                                     | Top-K sampling; -1 to disable            |\n| NVIDIA_NIM_MIN_P    ","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"","lvl3":""}},
+{"objectID":"c4d6bc20fbe9d189291d36b241fffab77fac1dcbe5ff66e1df5248840997c55d","title":"NVIDIA NIM Provider Guide","url":"/docs/getting-started/providers/nvidia-nim","content":"NVIDIA NIM Provider Guide\n\nHundreds of optimised AI models on NVIDIA's GPU-accelerated inference platform — or your own self-hosted NIM deployment\n\nOverview\n\nNVIDIA NIM (NVIDIA Inference Microservices) is a managed inference platform that hosts a large catalog of open-weight models — Meta Llama, DeepSeek, Mistral, Microsoft Phi, Google Gemma, and more — all GPU-optimised and served through an OpenAI-compatible API. You can also point NeuroLink at a self-hosted NIM cluster by overriding the base URL.\n\nKey Facts\nHosted base URL: https://integrate.api.nvidia.com/v1\nProtocol: OpenAI-compatible (/v1/chat/completions)\nVision: Yes, on supported models (Llama 3.2 Vision, etc.)\nReasoning: Yes, on Nemotron and DeepSeek-R1 variants\nStreaming: Supported\nTool calling: Supported on most models\nSelf-hosting: Override NVIDIA_NIM_BASE_URL to point at a private NIM cluster\n\nNIM-Specific Extras\n\nNIM supports additional generation parameters beyond the standard OpenAI surface: top_k, min_p, repetition_penalty, min_tokens, and per-model chat_template overrides. The NeuroLink provider automatically passes these via the providerOptions.openai.body mechanism. If a model rejects an unsupported parameter with HTTP 400, the provider retries the request with that parameter stripped.\n\nQuick Start\nGet an API Key\n\nSign up at https://build.nvidia.com and create an API key under API Keys.\nConfigure Environment\n\nAdd to your .env file:\nInstall NeuroLink\nGenerate Your First Response\n\nSupported Models\n\nNIM hosts hundreds of models. NeuroLink ships with these popular models pre-enumerated:\n\nOpenAI gpt-oss (default)\n\n| Model ID             | Context | Vision | Reasoning |\n| -------------------- | ------- | ------ | --------- |\n| openai/gpt-oss-20b | 128K    | No     | No        |\n\nMeta Llama\n\n| Model ID                             | Context | Vision | Reasoning |\n| ------------------------------------ | ------- | ------ | --------- |\n| meta/llama-3.3-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.1-405b-instruct       | 128K    | No     | No        |\n| meta/llama-3.1-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.2-90b-vision-instruct | 128K    | Yes    | No        |\n| meta/llama-3.2-11b-vision-instruct | 128K    | Yes    | No        |\n\nNVIDIA Nemotron (Reasoning)\n\n| Model ID                                 | Context | Vision | Reasoning |\n| ---------------------------------------- | ------- | ------ | --------- |\n| nvidia/llama-3.3-nemotron-super-49b-v1 | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-nano-8b-v1   | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-70b-instruct | 128K    | No     | Yes       |\n\nDeepSeek (Hosted on NIM)\n\n| Model ID                                    | Context | Vision | Reasoning |\n| ------------------------------------------- | ------- | ------ | --------- |\n| deepseek-ai/deepseek-r1                   | 128K    | No     | Yes       |\n| deepseek-ai/deepseek-r1-distill-llama-70b | 128K    | No     | Yes       |\n\nOther Models\n\n| Model ID                                | Context | Notes            |\n| --------------------------------------- | ------- | ---------------- |\n| mistralai/mixtral-8x22b-instruct-v0.1 | 64K     | Large MoE        |\n| mistralai/mixtral-8x7b-instruct-v0.1  | 32K     | Efficient MoE    |\n| microsoft/phi-4                       | 16K     | Compact, capable |\n| google/gemma-3-27b-it                 | 128K    | Google Gemma     |\n\nAvailability, checked 2026-10-02. NVIDIA's GET https://integrate.api.nvidia.com/v1/models did not list meta/llama-3.3-70b-instruct, meta/llama-3.1-70b-instruct, deepseek-ai/deepseek-r1-distill-llama-70b or google/gemma-3-27b-it. The NvidiaNimModels enum records a probe on 2026-08-26 in which these answered \"no longer available\", and keeps the members so existing code compiles. meta/llama-3.1-405b-instruct, deepseek-ai/deepseek-r1, mistralai/mixtral-8x22b-instruct-v0.1, mistralai/mixtral-8x7b-instruct-v0.1 and microsoft/phi-4 were also absent from that list on that date, and meta/llama-3.2-90b-vision-instruct was listed.\n\nBrowse the full catalog at https://build.nvidia.com/models. You can pass any model ID via --model or model: — NIM returns 404 for IDs that are not in the catalog.\n\nSDK Usage\n\nBasic Generation\n\nUsing a Specific Model\n\nReasoning with thinkingLevel\n\nReasoning-capable models (Nemotron, DeepSeek-R1) accept a thinking flag via NIM's chat_template_kwargs. Pass thinkingLevel to activate it:\n\nLevels: minimal (no thinking) | low | medium | high\n\nIf the model does not support chat_template_kwargs.thinking, the provider automatically retries without it.\n\nStreaming\n\nPer-Call Credential Override\n\nFor self-hosted NIM clusters, override the base URL per call:\n\nCLI Usage\n\nBasic Commands\n\nProvider Aliases\n\n| Alias        | Example                 |\n| ------------ | ----------------------- |\n| nvidia-nim | --provider nvidia-nim |\n| nvidia     | --provider nvidia     |\n| nim        | --provider nim        |\n\nC","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"","lvl3":""}},
 {"objectID":"b3c7e6b9de265c001fb3ea1118f2467719f7d5d7b47d0d1219803161adaa97f3","title":"NVIDIA NIM Provider Guide","url":"/docs/getting-started/providers/nvidia-nim#nvidia-nim-provider-guide","content":"Hundreds of optimised AI models on NVIDIA's GPU-accelerated inference platform — or your own self-hosted NIM deployment","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"NVIDIA NIM Provider Guide","lvl3":""}},
 {"objectID":"bdd47b6d8dc71c8c6e099e6aaf6d4b1322ee018e437203f0f8213316ead4bc6a","title":"Overview","url":"/docs/getting-started/providers/nvidia-nim#overview","content":"NVIDIA NIM (NVIDIA Inference Microservices) is a managed inference platform that hosts a large catalog of open-weight models — Meta Llama, DeepSeek, Mistral, Microsoft Phi, Google Gemma, and more — all GPU-optimised and served through an OpenAI-compatible API. You can also point NeuroLink at a self-hosted NIM cluster by overriding the base URL.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Overview","lvl3":""}},
 {"objectID":"70e960324b4417bd885aed19e36a44abbedba28c263ff2bf2e6024c995e0915d","title":"Key Facts","url":"/docs/getting-started/providers/nvidia-nim#key-facts","content":"Hosted base URL: https://integrate.api.nvidia.com/v1\nProtocol: OpenAI-compatible (/v1/chat/completions)\nVision: Yes, on supported models (Llama 3.2 Vision, etc.)\nReasoning: Yes, on Nemotron and DeepSeek-R1 variants\nStreaming: Supported\nTool calling: Supported on most models\nSelf-hosting: Override NVIDIA_NIM_BASE_URL to point at a private NIM cluster","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Key Facts","lvl3":""}},
@@ -6572,16 +6572,17 @@
 {"objectID":"fef841a51c5f7d0ab1473cf9655493c106b37a29d1c362b0bf6a1f309fbfeada","title":"1. Get an API Key","url":"/docs/getting-started/providers/nvidia-nim#1-get-an-api-key","content":"Sign up at https://build.nvidia.com and create an API key under API Keys.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"1. Get an API Key","lvl3":""}},
 {"objectID":"54f576f3812e2af49eeaac655066e1c3b52337b017abcb2e7ba56befc2f8b075","title":"2. Configure Environment","url":"/docs/getting-started/providers/nvidia-nim#2-configure-environment","content":"Add to your .env file:\n\n`bash","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"2. Configure Environment","lvl3":""}},
 {"objectID":"af9aefd38ecb7156515e45a9b14f49b2d4e35f5103766a34ee526c7376581718","title":"Required","url":"/docs/getting-started/providers/nvidia-nim#required","content":"NVIDIANIMAPI_KEY=nvapi-...","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Required","lvl3":""}},
-{"objectID":"d477283d648f6354308e128d5720c69502861661f47aaa21c75b9bfc8df19dea","title":"Optional: override the default model (default: meta/llama-3.3-70b-instruct)","url":"/docs/getting-started/providers/nvidia-nim#optional-override-the-default-model-default-metallama-33-70b-instruct","content":"NVIDIANIMMODEL=meta/llama-3.3-70b-instruct","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Optional: override the default model (default: meta/llama-3.3-70b-instruct)","lvl3":""}},
+{"objectID":"90754e9183ad1fcb9cf9ff0bd3c6940d17bc19cd5ed4f3ff2ff789df5ca8d154","title":"Optional: override the default model (default: openai/gpt-oss-20b)","url":"/docs/getting-started/providers/nvidia-nim#optional-override-the-default-model-default-openaigpt-oss-20b","content":"NVIDIANIMMODEL=openai/gpt-oss-20b","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Optional: override the default model (default: openai/gpt-oss-20b)","lvl3":""}},
 {"objectID":"b76f23a2ce32b0caf9a68d600d71065bec032c30b4c4ec20e1aeb3ab10bc34f2","title":"Optional: self-hosted NIM base URL (default: https://integrate.api.nvidia.com/v1)","url":"/docs/getting-started/providers/nvidia-nim#optional-self-hosted-nim-base-url-default-httpsintegrateapinvidiacomv1","content":"NVIDIANIMBASE_URL=https://integrate.api.nvidia.com/v1","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Optional: self-hosted NIM base URL (default: https://integrate.api.nvidia.com/v1)","lvl3":""}},
 {"objectID":"4d821a39a9f5c5c88e459de816567ecbbcf3f8a9166888d78b326669c61c8803","title":"Optional: NIM-specific generation parameters","url":"/docs/getting-started/providers/nvidia-nim#optional-nim-specific-generation-parameters","content":"NVIDIANIMTOP_K=40\nNVIDIANIMMIN_P=0.05\nNVIDIANIMREPETITION_PENALTY=1.1\nNVIDIANIMMIN_TOKENS=1\nNVIDIANIMCHAT_TEMPLATE=\n`","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Optional: NIM-specific generation parameters","lvl3":""}},
 {"objectID":"2c10c41d2143a2e957aca12ab30393841715f915cb586f80c76ceb54ca1f5dd5","title":"3. Install NeuroLink","url":"/docs/getting-started/providers/nvidia-nim#3-install-neurolink","content":"`bash\nnpm install @juspay/neurolink","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"3. Install NeuroLink","lvl3":""}},
 {"objectID":"cc6f7b1b64201da0c3ccaa840b81bfddec77ed18571c72532e43631cb84467ff","title":"or","url":"/docs/getting-started/providers/nvidia-nim#or","content":"pnpm add @juspay/neurolink\n`","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"or","lvl3":""}},
 {"objectID":"bc1d7b9f9cd1f0093bceb3e9322d3a8907621c41375b95d6991236ad35217786","title":"Supported Models","url":"/docs/getting-started/providers/nvidia-nim#supported-models","content":"NIM hosts hundreds of models. NeuroLink ships with these popular models pre-enumerated:","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Supported Models","lvl3":""}},
+{"objectID":"e0e7cb23f65e12d5e482c0dc5315010163c449defc6476af5e56bec0f2c67400","title":"OpenAI gpt-oss (default)","url":"/docs/getting-started/providers/nvidia-nim#openai-gpt-oss-default","content":"| Model ID             | Context | Vision | Reasoning |\n| -------------------- | ------- | ------ | --------- |\n| openai/gpt-oss-20b | 128K    | No     | No        |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"OpenAI gpt-oss (default)","lvl3":""}},
 {"objectID":"041d61fa62afb3ac253f977f945401de38379d0f20f40075103a0fa368137747","title":"Meta Llama","url":"/docs/getting-started/providers/nvidia-nim#meta-llama","content":"| Model ID                             | Context | Vision | Reasoning |\n| ------------------------------------ | ------- | ------ | --------- |\n| meta/llama-3.3-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.1-405b-instruct       | 128K    | No     | No        |\n| meta/llama-3.1-70b-instruct        | 128K    | No     | No        |\n| meta/llama-3.2-90b-vision-instruct | 128K    | Yes    | No        |\n| meta/llama-3.2-11b-vision-instruct | 128K    | Yes    | No        |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Meta Llama","lvl3":""}},
 {"objectID":"3c188653cca3eebd5cc90f43252ad42029a4fa888bfb2b470139e4e0db08e26a","title":"NVIDIA Nemotron (Reasoning)","url":"/docs/getting-started/providers/nvidia-nim#nvidia-nemotron-reasoning","content":"| Model ID                                 | Context | Vision | Reasoning |\n| ---------------------------------------- | ------- | ------ | --------- |\n| nvidia/llama-3.3-nemotron-super-49b-v1 | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-nano-8b-v1   | 128K    | No     | Yes       |\n| nvidia/llama-3.1-nemotron-70b-instruct | 128K    | No     | Yes       |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"NVIDIA Nemotron (Reasoning)","lvl3":""}},
 {"objectID":"4d9aca0894a65a52648a73478e1c73b487411326cf63e05eb8159fd8fef0c5aa","title":"DeepSeek (Hosted on NIM)","url":"/docs/getting-started/providers/nvidia-nim#deepseek-hosted-on-nim","content":"| Model ID                                    | Context | Vision | Reasoning |\n| ------------------------------------------- | ------- | ------ | --------- |\n| deepseek-ai/deepseek-r1                   | 128K    | No     | Yes       |\n| deepseek-ai/deepseek-r1-distill-llama-70b | 128K    | No     | Yes       |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"DeepSeek (Hosted on NIM)","lvl3":""}},
-{"objectID":"8eccc475db8aa2e29bd49cd7c0ecb93be5b813979af07749f0c218abe7a71bd8","title":"Other Models","url":"/docs/getting-started/providers/nvidia-nim#other-models","content":"| Model ID                                | Context | Notes            |\n| --------------------------------------- | ------- | ---------------- |\n| mistralai/mixtral-8x22b-instruct-v0.1 | 64K     | Large MoE        |\n| mistralai/mixtral-8x7b-instruct-v0.1  | 32K     | Efficient MoE    |\n| microsoft/phi-4                       | 16K     | Compact, capable |\n| google/gemma-3-27b-it                 | 128K    | Google Gemma     |\n\nBrowse the full catalog at https://build.nvidia.com/models. You can pass any model ID via --model or model: — NIM returns 404 for IDs that are not in the catalog.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Other Models","lvl3":""}},
+{"objectID":"8eccc475db8aa2e29bd49cd7c0ecb93be5b813979af07749f0c218abe7a71bd8","title":"Other Models","url":"/docs/getting-started/providers/nvidia-nim#other-models","content":"| Model ID                                | Context | Notes            |\n| --------------------------------------- | ------- | ---------------- |\n| mistralai/mixtral-8x22b-instruct-v0.1 | 64K     | Large MoE        |\n| mistralai/mixtral-8x7b-instruct-v0.1  | 32K     | Efficient MoE    |\n| microsoft/phi-4                       | 16K     | Compact, capable |\n| google/gemma-3-27b-it                 | 128K    | Google Gemma     |\n\nAvailability, checked 2026-10-02. NVIDIA's GET https://integrate.api.nvidia.com/v1/models did not list meta/llama-3.3-70b-instruct, meta/llama-3.1-70b-instruct, deepseek-ai/deepseek-r1-distill-llama-70b or google/gemma-3-27b-it. The NvidiaNimModels enum records a probe on 2026-08-26 in which these answered \"no longer available\", and keeps the members so existing code compiles. meta/llama-3.1-405b-instruct, deepseek-ai/deepseek-r1, mistralai/mixtral-8x22b-instruct-v0.1, mistralai/mixtral-8x7b-instruct-v0.1 and microsoft/phi-4 were also absent from that list on that date, and meta/llama-3.2-90b-vision-instruct was listed.\n\nBrowse the full catalog at https://build.nvidia.com/models. You can pass any model ID via --model or model: — NIM returns 404 for IDs that are not in the catalog.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Other Models","lvl3":""}},
 {"objectID":"def6f89a4845940b499d11b73fb1b4c4aee4bf4315cfbf3828168d4b874c3f39","title":"Reasoning with thinkingLevel","url":"/docs/getting-started/providers/nvidia-nim#reasoning-with-thinkinglevel","content":"Reasoning-capable models (Nemotron, DeepSeek-R1) accept a thinking flag via NIM's chat_template_kwargs. Pass thinkingLevel to activate it:\n\nLevels: minimal (no thinking) | low | medium | high\n\nIf the model does not support chat_template_kwargs.thinking, the provider automatically retries without it.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Reasoning with thinkingLevel","lvl3":""}},
 {"objectID":"7cf16919f6a4e4931c46ea94384eb786a1c93d1b73d3086b3247bb826f5844cb","title":"Per-Call Credential Override","url":"/docs/getting-started/providers/nvidia-nim#per-call-credential-override","content":"For self-hosted NIM clusters, override the base URL per call:","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Per-Call Credential Override","lvl3":""}},
 {"objectID":"d997142b7549ed2c6ac3cb1c791f610155830be4d3dfeacd34fc36fed024fcba","title":"Basic Commands","url":"/docs/getting-started/providers/nvidia-nim#basic-commands","content":"`bash","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Basic Commands","lvl3":""}},
@@ -6591,13 +6592,13 @@
 {"objectID":"a5fc082bd08fdd0492c3bbe3162c63cbeb7207a09a8258824eda8872e8de4232","title":"Reasoning model with thinking enabled","url":"/docs/getting-started/providers/nvidia-nim#reasoning-model-with-thinking-enabled","content":"pnpm run cli generate \"Solve: what is 17! mod 13?\" \\\n  --provider nvidia-nim \\\n  --model deepseek-ai/deepseek-r1 \\\n  --thinking-level high","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Reasoning model with thinking enabled","lvl3":""}},
 {"objectID":"58d637c0f9375a930ef480da8d002cb98dab13c49a1a23a555a16a5ead6c7bd1","title":"Interactive loop","url":"/docs/getting-started/providers/nvidia-nim#interactive-loop","content":"pnpm run cli loop --provider nvidia-nim\n`","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Interactive loop","lvl3":""}},
 {"objectID":"6d2c2771b00915a5ce976d69a16c5a8a23bd35f7efd262e21884530d20c0a0e4","title":"Provider Aliases","url":"/docs/getting-started/providers/nvidia-nim#provider-aliases","content":"| Alias        | Example                 |\n| ------------ | ----------------------- |\n| nvidia-nim | --provider nvidia-nim |\n| nvidia     | --provider nvidia     |\n| nim        | --provider nim        |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Provider Aliases","lvl3":""}},
-{"objectID":"7fe825d689bae105fb7cc49ed299fd96c65231ea4ad501fa3c1c3ed36df332e7","title":"Configuration Reference","url":"/docs/getting-started/providers/nvidia-nim#configuration-reference","content":"| Environment Variable            | Required | Default                               | Description                                |\n| ------------------------------- | -------- | ------------------------------------- | ------------------------------------------ |\n| NVIDIA_NIM_API_KEY            | Yes      | —                                     | NVIDIA NIM API key (starts with nvapi-)  |\n| NVIDIA_NIM_MODEL              | No       | meta/llama-3.3-70b-instruct         | Default model                              |\n| NVIDIA_NIM_BASE_URL           | No       | https://integrate.api.nvidia.com/v1 | Base URL (override for self-hosted NIM)    |\n| NVIDIA_NIM_TOP_K              | No       | —                                     | Top-K sampling; -1 to disable            |\n| NVIDIA_NIM_MIN_P              | No       | —                                     | Minimum token probability; 0 to disable  |\n| NVIDIA_NIM_REPETITION_PENALTY | No       | —                                     | Anti-repetition factor; 1 is neutral     |\n| NVIDIA_NIM_MIN_TOKENS         | No       | —                                     | Minimum output length in tokens            |\n| NVIDIA_NIM_CHAT_TEMPLATE      | No       | —                                     | Override the model's default chat template |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Configuration Reference","lvl3":""}},
+{"objectID":"7fe825d689bae105fb7cc49ed299fd96c65231ea4ad501fa3c1c3ed36df332e7","title":"Configuration Reference","url":"/docs/getting-started/providers/nvidia-nim#configuration-reference","content":"| Environment Variable            | Required | Default                               | Description                                |\n| ------------------------------- | -------- | ------------------------------------- | ------------------------------------------ |\n| NVIDIA_NIM_API_KEY            | Yes      | —                                     | NVIDIA NIM API key (starts with nvapi-)  |\n| NVIDIA_NIM_MODEL              | No       | openai/gpt-oss-20b                  | Default model                              |\n| NVIDIA_NIM_BASE_URL           | No       | https://integrate.api.nvidia.com/v1 | Base URL (override for self-hosted NIM)    |\n| NVIDIA_NIM_TOP_K              | No       | —                                     | Top-K sampling; -1 to disable            |\n| NVIDIA_NIM_MIN_P              | No       | —                                     | Minimum token probability; 0 to disable  |\n| NVIDIA_NIM_REPETITION_PENALTY | No       | —                                     | Anti-repetition factor; 1 is neutral     |\n| NVIDIA_NIM_MIN_TOKENS         | No       | —                                     | Minimum output length in tokens            |\n| NVIDIA_NIM_CHAT_TEMPLATE      | No       | —                                     | Override the model's default chat template |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Configuration Reference","lvl3":""}},
 {"objectID":"56a3e36fceef9780446d157f62f778d26baa8a79c25571830cc4d8ca8d7b180e","title":"Self-Hosted NIM","url":"/docs/getting-started/providers/nvidia-nim#self-hosted-nim","content":"If you run NIM on your own GPU cluster, set NVIDIA_NIM_BASE_URL to point at your cluster. Authentication is still forwarded via Authorization: Bearer, so set NVIDIA_NIM_API_KEY to any non-empty value if your cluster does not require it (or to your actual cluster token if it does).","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Self-Hosted NIM","lvl3":""}},
 {"objectID":"05c1f3aa67afc6e69a07d0a382fc1433f4f4bc1b1aaeba20f0dc2dbd64b4da43","title":"Feature Support","url":"/docs/getting-started/providers/nvidia-nim#feature-support","content":"| Feature         | Supported | Notes                                                 |\n| --------------- | --------- | ----------------------------------------------------- |\n| Text generation | Yes       |                                                       |\n| Streaming       | Yes       |                                                       |\n| Tool calling    | Yes       | Most models; depends on model support                 |\n| Vision / images | Yes       | Model-dependent (Llama 3.2 Vision, etc.)              |\n| Reasoning trace | Yes       | Nemotron and DeepSeek-R1 variants via thinkingLevel |\n| Embeddings      | No        | Use OpenAI or Bedrock for embeddings                  |","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Feature Support","lvl3":""}},
 {"objectID":"e113acee1e904cde2e3295ebbb99df56e1d4d712cf149f774c8fbaad6ada6770","title":"\"Invalid NVIDIA NIM API key\"","url":"/docs/getting-started/providers/nvidia-nim#invalid-nvidia-nim-api-key","content":"The NVIDIA_NIM_API_KEY is missing, expired, or incorrect.\n\nGet or rotate keys at https://build.nvidia.com/settings/api-keys.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"\"Invalid NVIDIA NIM API key\"","lvl3":""}},
 {"objectID":"25669258e3b272543a86496e295868082d0146aff134a84ead2bdf65fccb3ffb","title":"\"NVIDIA NIM rate limit exceeded\"","url":"/docs/getting-started/providers/nvidia-nim#nvidia-nim-rate-limit-exceeded","content":"Your account has hit its request-per-minute or token-per-day limit. Upgrade your account, reduce request frequency, or implement backoff. Check your current usage at https://build.nvidia.com/usage.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"\"NVIDIA NIM rate limit exceeded\"","lvl3":""}},
 {"objectID":"8fa05f7794655a8b226d68b36d26031af67d19af32b9435a02357b7f8e719862","title":"\"NVIDIA NIM model not available\"","url":"/docs/getting-started/providers/nvidia-nim#nvidia-nim-model-not-available","content":"The model ID is not in the NIM catalog, or your account tier does not have access.\n\n`bash","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"\"NVIDIA NIM model not available\"","lvl3":""}},
-{"objectID":"c1e90df7e7b48c54aeed13aae5ceb77c009f7c3d69d3f000c4253e652f8a29d6","title":"Browse the catalog","url":"/docs/getting-started/providers/nvidia-nim#browse-the-catalog","content":"open https://build.nvidia.com/models\n\n\nUse the exact model ID shown on the model's page (e.g., meta/llama-3.3-70b-instruct`).","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Browse the catalog","lvl3":""}},
+{"objectID":"c1e90df7e7b48c54aeed13aae5ceb77c009f7c3d69d3f000c4253e652f8a29d6","title":"Browse the catalog","url":"/docs/getting-started/providers/nvidia-nim#browse-the-catalog","content":"open https://build.nvidia.com/models\n\n\nUse the exact model ID shown on the model's page (e.g., openai/gpt-oss-20b`).","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Browse the catalog","lvl3":""}},
 {"objectID":"34d171d7ade8a31dfc31588bfccef6e920cd37a662a53089aabb5983a7d2694c","title":"\"NVIDIA NIM quota exceeded\"","url":"/docs/getting-started/providers/nvidia-nim#nvidia-nim-quota-exceeded","content":"Account-level token or compute quota reached. Check your NIM dashboard.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"\"NVIDIA NIM quota exceeded\"","lvl3":""}},
 {"objectID":"0697a2c79b09552d2abfdc5d788f32dfeec83add6b8a40e67389ef43b6f77ac9","title":"HTTP 400 with reasoning_budget or chat_template in the error","url":"/docs/getting-started/providers/nvidia-nim#http-400-with-reasoning_budget-or-chat_template-in-the-error","content":"The model does not support one of the NIM-specific extras. The provider automatically retries without the rejected parameter. If you see this error surfaced, it means the second attempt also failed — check the rest of the error message for the root cause.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"HTTP 400 with reasoning_budget or chat_template in the error","lvl3":""}},
 {"objectID":"403e25d3806bf5242a71a8e510fac16e068f907380e7f22d15d2393cbf3f28f4","title":"Thinking level has no visible effect","url":"/docs/getting-started/providers/nvidia-nim#thinking-level-has-no-visible-effect","content":"Not all models support chat_template_kwargs.thinking. If the model rejects the parameter, the provider retries the request without it and produces a normal (non-reasoning) response. Use a Nemotron or DeepSeek-R1 model for guaranteed reasoning support.","hierarchy":{"lvl0":"Getting Started","lvl1":"NVIDIA NIM Provider Guide","lvl2":"Thinking level has no visible effect","lvl3":""}},
@@ -6862,13 +6863,13 @@
 {"objectID":"e4a90009c7bcd26ccb6a5309dfd66d6dc17c46b9abbfb292b212fe38b41c1b98","title":"Verification status","url":"/docs/getting-started/providers/parasail#verification-status","content":"Tier-2 onboarding requires evidence before a provider is accepted, and\npnpm run verify:provider-onboarding gates it in CI. This is what the\ncatalog currently records for Parasail — docs-verified only, not yet\nlive-verified:\n\n| Probe                     | Result                                                                                                                                                                                                                                                                                                                                                                                                    |\n| ------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| Roster                    | unauthenticated GET https://api.parasail.io/v1/models answered HTTP 401 on 2026-09-29 — no key used, so the model ids are taken from the Models page (retrieved 2026-09-29) and the roster is not verified                                                                                                                                                                                              |\n| Billing                   | the Pricing page says \"To use models, create API keys, and access paid platform services, add a credit card. Without a card, you can still view the platform but can't use models or API keys.\"; the Limits and Quotas page has a \"Serverless - Free\" rate-limit row; no signup was performed, 2026-09-29                                                                                                 |\n| Auth-failure shape        | unauthenticated GETs of /v1/models, /v1/cha","hierarchy":{"lvl0":"Getting Started","lvl1":"Parasail Provider Guide","lvl2":"Verification status","lvl3":""}},
 {"objectID":"7524ca685dabe987464872496673f6a1795cb28e182c63b7eea6e8e6f2f7de4d","title":"Documented error statuses","url":"/docs/getting-started/providers/parasail#documented-error-statuses","content":"The\nChat Completions reference\nsays \"Errors use OpenAI-compatible status codes and a JSON body with an error\nobject (message, type, code).\" and lists these statuses:\n\n| Status | Parasail's description                                                                 |\n| ------ | -------------------------------------------------------------------------------------- |\n| 400  | Malformed request—invalid JSON, missing required field, or unsupported parameter value |\n| 401  | Missing or invalid API key                                                             |\n| 403  | API key lacks access to the requested model or resource                                |\n| 404  | Model or endpoint not found                                                            |\n| 429  | Rate limit or quota exceeded—back off and retry                                        |\n| 500  | Internal server error—retry with exponential backoff                                   |\n\nThe\nTool/Function Calling guide\nsays of models outside its table: \"Depending on the model, a request with tools\nreturns an error or plain text with no tool_calls.\"","hierarchy":{"lvl0":"Getting Started","lvl1":"Parasail Provider Guide","lvl2":"Documented error statuses","lvl3":""}},
 {"objectID":"f48bba8c2e9813155262f99a41ecc882205dc495aa27c27bb0b50c527a0f8a5d","title":"See also","url":"/docs/getting-started/providers/parasail#see-also","content":"Provider setup overview\nAll providers\nTier-2 onboarding — how this provider's JSON becomes a working integration\nProvider feature compatibility\nParasail docs pages used for this entry: Models, Pricing, Authentication, Chat Completions reference, Tool/Function Calling, Structured Output, Multi-Modal, Model-specific Notes, Limits and Quotas","hierarchy":{"lvl0":"Getting Started","lvl1":"Parasail Provider Guide","lvl2":"See also","lvl3":""}},
-{"objectID":"5e3a315038085bf3d2d31cbaaf7a07c7df8228093af856dd3a697f19736edf24","title":"Pareto Inference Provider Guide","url":"/docs/getting-started/providers/pareto-inference","content":"Pareto Inference Provider Guide\n\nVerification status: this entry is docs- and roster-verified only —\nnot yet live-verified. It was onboarded credential-free: no Pareto\nInference API key was ever created or used. Every field below comes from\nPareto's public docs (https://docs.paretoinference.com/) and an\nunauthenticated GET /v1/models call (no auth header). No POST\nrequest was ever sent, so no generate, stream, tool-calling or\nstructured-output capability has been exercised end to end. Treat\ncapabilities.tools and .thinking as documentation-grade, not\nwire-proven, until evidence.liveMatrix is filled in by a live run.\n\nPareto Inference is a Tier-2 catalog provider: OpenAI-wire-compatible\nwith no behavioural quirks, so its entire integration is one JSON file\n(src/lib/providers/catalog/pareto-inference.json) rather than hand-written\ncode. That file is the source of truth for everything on this page.\n\nKey Facts\nProvider id: pareto-inference\nProtocol: OpenAI-compatible (/chat/completions) — \"Pareto accepts\n  only OpenAI Chat Completions.\"\nBase URL: https://api.paretoinference.com/v1\nDefault model: z-ai/glm-5.3-flash (GLM 5.3 Flash; the docs also\n  document a bare alias, glm-5.3-flash, for the same model)\nModels in catalog: 1 — the only model Pareto serves. Confirmed by an\n  unauthenticated GET /v1/models (no key needed), which returns exactly\n  one id.\nStreaming: documented (SSE, text in choices[0].delta.content, tool\n  calls in choices[0].delta.tool_calls, terminal data: [DONE]) — not\n  live-probed.\nTool calling: documented (tools, tool_choice) but the vendor's own\n  request-parameter table hedges it with \"Model support can differ,\" with no\n  model-specific confirmation — set to model-dependent, not true.\nStructured output: response_format is a documented parameter but\n  carries the same \"Model support can differ\" hedge as tools; the catalog\n  schema has no model-dependent option for this field, so it is set\n  false rather than asserted from a hedged table. structuredOutputWithTools\n  is false — no combined probe was possible credential-free.\nReasoning: reasoning_effort (\"Sets the reasoning level\") is\n  documented as a plain, unhedged request parameter — unlike tools /\n  response_format / tool_choice / logprobs, it carries no \"Model\n  support can differ\" caveat — the basis for capabilities.thinking: true.\nVision: not documented anywhere on the site.\nEmbeddings: not documented anywhere on the site.\nBilling: no free tier. A payment card and prepaid credits are\n  required before a key is issued.\nSignup: https://paretoinference.com/dashboard\nKey format: none declared.\nData handling: Pareto documents zero data retention (ZDR) — request\n  content is not kept after a request finishes.\n\nQuick Start\nGet an API key\nVisit: https://paretoinference.com/dashboard and sign in\nAdd a payment card — Pareto has no free tier; a card and prepaid credits\n   are required before a key is issued\n   (https://docs.paretoinference.com/pricing)\nBuy prepaid credits, then create your API key from the dashboard\nSet PARETO_INFERENCE_API_KEY in your .env file\nConfigure\nUse it\n\nPer-request credentials work as they do for every provider:\n\nModels\n\n| Model                   | Context   | Vision | $/M in · out · cached         | Notes                                                                                                                                                                                                                                       |\n| ----------------------- | --------- | ------ | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| z-ai/glm-5.3-flash ⭐ | 131,072\\ | no     | $0.03 / $0.10 (cached $0.006) | Only model Pareto serves. Max output documented as 1–131,072 tokens (default 131,072). \\Context window is not published by Pareto — this figure reuses the documented output ceiling as a conservative floor, not a stated context length. |\n\n\\ Pareto's own docs state it does not* disclose GLM 5.3 Flash's serving\ndetails (\"We do not disclose the technical details of our GLM 5.3 Flash\nserving system\"), and no context-window number appears on any public page.\nmodels.defaultContextWindow (131,072) is therefore the documented\noutput ceiling reused as a floor, flagged in the catalog entry's\ndescription — correct it once Pareto publishes a real figure or an\nauthenticated probe reveals one.\n\nFallback order: none — z-ai/glm-5.3-flash is the only model Pareto\nserves, so it is both the default and its own (schema-exempt) single-entry\nfallback list.\n\nVerification status\n\nTier-2 onboarding requires evidence before a provider is accepted, and\npnpm run verify:provider-onboarding gates it in CI. This is what the\ncatalog records for Pareto Inference — and, just as importantly, what it\ndoes not yet record:\n\n| Probe           ","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"","lvl3":""}},
+{"objectID":"5e3a315038085bf3d2d31cbaaf7a07c7df8228093af856dd3a697f19736edf24","title":"Pareto Inference Provider Guide","url":"/docs/getting-started/providers/pareto-inference","content":"Pareto Inference Provider Guide\n\nVerification status: this entry is docs- and roster-verified only —\nnot yet live-verified. It was onboarded credential-free: no Pareto\nInference API key was ever created or used. Every field below comes from\nPareto's public docs (https://docs.paretoinference.com/) and an\nunauthenticated GET /v1/models call (no auth header). No POST\nrequest was ever sent, so no generate, stream, tool-calling or\nstructured-output capability has been exercised end to end. Treat\ncapabilities.tools and .thinking as documentation-grade, not\nwire-proven, until evidence.liveMatrix is filled in by a live run.\n\nPareto Inference is a Tier-2 catalog provider: OpenAI-wire-compatible\nwith no behavioural quirks, so its entire integration is one JSON file\n(src/lib/providers/catalog/pareto-inference.json) rather than hand-written\ncode. That file is the source of truth for everything on this page.\n\nKey Facts\nProvider id: pareto-inference\nProtocol: OpenAI-compatible (/chat/completions) — \"Pareto accepts\n  only OpenAI Chat Completions.\"\nBase URL: https://api.paretoinference.com/v1\nDefault model: z-ai/glm-5.3-flash (GLM 5.3 Flash; the docs also\n  document a bare alias, glm-5.3-flash, for the same model)\nModels in catalog: 1 — the only model Pareto serves. Confirmed by an\n  unauthenticated GET /v1/models (no key needed), which returns exactly\n  one id.\nStreaming: documented (SSE, text in choices[0].delta.content, tool\n  calls in choices[0].delta.tool_calls, terminal data: [DONE]) — not\n  live-probed.\nTool calling: documented (tools, tool_choice) but the vendor's own\n  request-parameter table hedges it with \"Model support can differ,\" with no\n  model-specific confirmation — set to model-dependent, not true.\nStructured output: response_format is a documented parameter but\n  carries the same \"Model support can differ\" hedge as tools; the catalog\n  schema has no model-dependent option for this field, so it is set\n  false rather than asserted from a hedged table. structuredOutputWithTools\n  is false — no combined probe was possible credential-free.\nReasoning: reasoning_effort (\"Sets the reasoning level\") is\n  documented as a plain, unhedged request parameter — unlike tools /\n  response_format / tool_choice / logprobs, it carries no \"Model\n  support can differ\" caveat — the basis for capabilities.thinking: true.\nVision: not documented anywhere on the site.\nEmbeddings: not documented anywhere on the site.\nBilling: no free tier. A payment card and prepaid credits are\n  required before a key is issued.\nSignup: https://paretoinference.com/dashboard\nKey format: none declared.\nData handling: Pareto documents zero data retention (ZDR) — request\n  content is not kept after a request finishes.\n\nQuick Start\nGet an API key\nVisit: https://paretoinference.com/dashboard and sign in\nAdd a payment card — Pareto has no free tier; a card and prepaid credits\n   are required before a key is issued\n   (https://docs.paretoinference.com/pricing)\nBuy prepaid credits, then create your API key from the dashboard\nSet PARETO_INFERENCE_API_KEY in your .env file\nConfigure\nUse it\n\nPer-request credentials work as they do for every provider:\n\nModels\n\n| Model                   | Context   | Vision | $/M in · out · cached         | Notes                                                                                                                                                                                                                                       |\n| ----------------------- | --------- | ------ | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| z-ai/glm-5.3-flash ⭐ | 131,072\\ | no     | $0.09 / $0.30 (cached $0.018) | Only model Pareto serves. Max output documented as 1–131,072 tokens (default 131,072). \\Context window is not published by Pareto — this figure reuses the documented output ceiling as a conservative floor, not a stated context length. |\n\n\\ Pareto's own docs state it does not* disclose GLM 5.3 Flash's serving\ndetails (\"We do not disclose the technical details of our GLM 5.3 Flash\nserving system\"), and no context-window number appears on any public page.\nmodels.defaultContextWindow (131,072) is therefore the documented\noutput ceiling reused as a floor, flagged in the catalog entry's\ndescription — correct it once Pareto publishes a real figure or an\nauthenticated probe reveals one.\n\nFallback order: none — z-ai/glm-5.3-flash is the only model Pareto\nserves, so it is both the default and its own (schema-exempt) single-entry\nfallback list.\n\nVerification status\n\nTier-2 onboarding requires evidence before a provider is accepted, and\npnpm run verify:provider-onboarding gates it in CI. This is what the\ncatalog records for Pareto Inference — and, just as importantly, what it\ndoes not yet record:\n\n| Probe           ","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"","lvl3":""}},
 {"objectID":"6a8ee157b685405e8c480486655e4656ea272f63493e36006b7ac43e499b5e27","title":"Pareto Inference Provider Guide","url":"/docs/getting-started/providers/pareto-inference#pareto-inference-provider-guide","content":"Verification status: this entry is docs- and roster-verified only —\nnot yet live-verified. It was onboarded credential-free: no Pareto\nInference API key was ever created or used. Every field below comes from\nPareto's public docs (https://docs.paretoinference.com/) and an\nunauthenticated GET /v1/models call (no auth header). No POST\nrequest was ever sent, so no generate, stream, tool-calling or\nstructured-output capability has been exercised end to end. Treat\ncapabilities.tools and .thinking as documentation-grade, not\nwire-proven, until evidence.liveMatrix is filled in by a live run.\n\nPareto Inference is a Tier-2 catalog provider: OpenAI-wire-compatible\nwith no behavioural quirks, so its entire integration is one JSON file\n(src/lib/providers/catalog/pareto-inference.json) rather than hand-written\ncode. That file is the source of truth for everything on this page.","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Pareto Inference Provider Guide","lvl3":""}},
 {"objectID":"4098c852271191df4a2cf6cfcd19b7832504f8a2e019abcce21852d7bc292fe3","title":"Key Facts","url":"/docs/getting-started/providers/pareto-inference#key-facts","content":"Provider id: pareto-inference\nProtocol: OpenAI-compatible (/chat/completions) — \"Pareto accepts\n  only OpenAI Chat Completions.\"\nBase URL: https://api.paretoinference.com/v1\nDefault model: z-ai/glm-5.3-flash (GLM 5.3 Flash; the docs also\n  document a bare alias, glm-5.3-flash, for the same model)\nModels in catalog: 1 — the only model Pareto serves. Confirmed by an\n  unauthenticated GET /v1/models (no key needed), which returns exactly\n  one id.\nStreaming: documented (SSE, text in choices[0].delta.content, tool\n  calls in choices[0].delta.tool_calls, terminal data: [DONE]) — not\n  live-probed.\nTool calling: documented (tools, tool_choice) but the vendor's own\n  request-parameter table hedges it with \"Model support can differ,\" with no\n  model-specific confirmation — set to model-dependent, not true.\nStructured output: response_format is a documented parameter but\n  carries the same \"Model support can differ\" hedge as tools; the catalog\n  schema has no model-dependent option for this field, so it is set\n  false rather than asserted from a hedged table. structuredOutputWithTools\n  is false — no combined probe was possible credential-free.\nReasoning: reasoning_effort (\"Sets the reasoning level\") is\n  documented as a plain, unhedged request parameter — unlike tools /\n  response_format / tool_choice / logprobs, it carries no \"Model\n  support can differ\" caveat — the basis for capabilities.thinking: true.\nVision: not documented anywhere on the site.\nEmbeddings: not documented anywhere on the site.\nBilling: no free tier. A payment card and prepaid credits are\n  required before a key is issued.\nSignup: https://paretoinference.com/dashboard\nKey format: none declared.\nData handling: Pareto documents zero data retention (ZDR) — request\n  content is not kept after a request finishes.","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Key Facts","lvl3":""}},
 {"objectID":"caf0e714786bb6800e72192b4bab1eb0c3e5b66a58b7545323fa523c7af1fe96","title":"1. Get an API key","url":"/docs/getting-started/providers/pareto-inference#1-get-an-api-key","content":"Visit: https://paretoinference.com/dashboard and sign in\nAdd a payment card — Pareto has no free tier; a card and prepaid credits\n   are required before a key is issued\n   (https://docs.paretoinference.com/pricing)\nBuy prepaid credits, then create your API key from the dashboard\nSet PARETO_INFERENCE_API_KEY in your .env file","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"1. Get an API key","lvl3":""}},
 {"objectID":"6ebc7974cc2c829d61423aa76879f84bf39f510a89e5d69d413a837fe72d86c6","title":"3. Use it","url":"/docs/getting-started/providers/pareto-inference#3-use-it","content":"`bash","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"3. Use it","lvl3":""}},
 {"objectID":"88e47ec55c1f3481092432425a6fabfeda109649e1722eff1cc531f760df674c","title":"CLI","url":"/docs/getting-started/providers/pareto-inference#cli","content":"npx @juspay/neurolink generate \"Hello\" --provider pareto-inference\ntypescript\nawait neurolink.generate({\n  input: { text: \"Hello\" },\n  provider: \"pareto-inference\",\n  credentials: {\n    \"pareto-inference\": { apiKey: process.env.PARETOINFERENCEAPI_KEY },\n  },\n});\n`","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"CLI","lvl3":""}},
-{"objectID":"79c93948b6d461182578b28b94eb9fa1edbf0815941cdb6090462ac57a92a6ab","title":"Models","url":"/docs/getting-started/providers/pareto-inference#models","content":"| Model                   | Context   | Vision | $/M in · out · cached         | Notes                                                                                                                                                                                                                                       |\n| ----------------------- | --------- | ------ | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| z-ai/glm-5.3-flash ⭐ | 131,072\\ | no     | $0.03 / $0.10 (cached $0.006) | Only model Pareto serves. Max output documented as 1–131,072 tokens (default 131,072). \\Context window is not published by Pareto — this figure reuses the documented output ceiling as a conservative floor, not a stated context length. |\n\n\\ Pareto's own docs state it does not* disclose GLM 5.3 Flash's serving\ndetails (\"We do not disclose the technical details of our GLM 5.3 Flash\nserving system\"), and no context-window number appears on any public page.\nmodels.defaultContextWindow (131,072) is therefore the documented\noutput ceiling reused as a floor, flagged in the catalog entry's\ndescription — correct it once Pareto publishes a real figure or an\nauthenticated probe reveals one.\n\nFallback order: none — z-ai/glm-5.3-flash is the only model Pareto\nserves, so it is both the default and its own (schema-exempt) single-entry\nfallback list.","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Models","lvl3":""}},
+{"objectID":"79c93948b6d461182578b28b94eb9fa1edbf0815941cdb6090462ac57a92a6ab","title":"Models","url":"/docs/getting-started/providers/pareto-inference#models","content":"| Model                   | Context   | Vision | $/M in · out · cached         | Notes                                                                                                                                                                                                                                       |\n| ----------------------- | --------- | ------ | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| z-ai/glm-5.3-flash ⭐ | 131,072\\ | no     | $0.09 / $0.30 (cached $0.018) | Only model Pareto serves. Max output documented as 1–131,072 tokens (default 131,072). \\Context window is not published by Pareto — this figure reuses the documented output ceiling as a conservative floor, not a stated context length. |\n\n\\ Pareto's own docs state it does not* disclose GLM 5.3 Flash's serving\ndetails (\"We do not disclose the technical details of our GLM 5.3 Flash\nserving system\"), and no context-window number appears on any public page.\nmodels.defaultContextWindow (131,072) is therefore the documented\noutput ceiling reused as a floor, flagged in the catalog entry's\ndescription — correct it once Pareto publishes a real figure or an\nauthenticated probe reveals one.\n\nFallback order: none — z-ai/glm-5.3-flash is the only model Pareto\nserves, so it is both the default and its own (schema-exempt) single-entry\nfallback list.","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Models","lvl3":""}},
 {"objectID":"106d04b8baec018f872ce608c3b0dd33f8b3d43ada85072831a2e9ab9c64c46b","title":"Verification status","url":"/docs/getting-started/providers/pareto-inference#verification-status","content":"Tier-2 onboarding requires evidence before a provider is accepted, and\npnpm run verify:provider-onboarding gates it in CI. This is what the\ncatalog records for Pareto Inference — and, just as importantly, what it\ndoes not yet record:\n\n| Probe                 | Result                                                                                                                                                                                                                                                                                                                                                                                             |\n| --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| Roster                | unauthenticated GET /v1/models, HTTP 200, 1 model (z-ai/glm-5.3-flash), 2026-09-28. No API key was used or required for this call.                                                                                                                                                                                                                                                         |\n| Auth rejection        | Not probed. Would require a POST with an invalid key, which this credential-free onboarding pass does not send. Pareto's docs describe 401 as \"the chat API key is missing, invalid, or no longer active,\" with no example error body.                                                                                                                                                           |\n| Live capability sweep | Not run. evidence.liveMatrix is null. capabilities.tools and .thinking are set from the v","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Verification status","lvl3":""}},
 {"objectID":"b844edaf09c93fb3089bbf0a71254c5039dedf111e2b59563321484496862aa4","title":"Troubleshooting","url":"/docs/getting-started/providers/pareto-inference#troubleshooting","content":"| Symptom                                 | Cause                                                                | Fix                                                                                                                             |\n| --------------------------------------- | -------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- |\n| Invalid Pareto Inference API key      | PARETO_INFERENCE_API_KEY unset, wrong, or expired                  | Pareto documents 401 as \"the chat API key is missing, invalid, or no longer active\" — check or replace the key in the dashboard |\n| 429 with credit_exhausted             | Prepaid balance is empty                                             | Buy more prepaid credits                                                                                                        |\n| 429 with credit_insufficient          | Held credits (reserved for max_tokens) are below what's needed     | Lower max_tokens, or buy more credits                                                                                         |\n| 429 with model_capacity               | The model itself is at capacity — not an account-level limit         | Honor the Retry-After header and retry                                                                                        |\n| 502 mid-stream                          | The model request failed after the stream started                    | Retry the request; check whether a tool call already ran before retrying                                                        |\n| 503                                     | Temporary Pareto-side failure (e.g. maintenance, GPU outage)         | Wait and retry; Retry-After gives the delay in seconds                                                                        |\n| No context-window figure in the catalog | Pareto does no","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"Troubleshooting","lvl3":""}},
 {"objectID":"57b8457eff592f229d3c0f0fb7c73148341ecffd72bb29c15d73f1471b1fd886","title":"See also","url":"/docs/getting-started/providers/pareto-inference#see-also","content":"Provider setup overview\nAll providers\nTier-2 onboarding — how this provider's JSON becomes a working integration\nProvider feature compatibility","hierarchy":{"lvl0":"Getting Started","lvl1":"Pareto Inference Provider Guide","lvl2":"See also","lvl3":""}},

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@juspay/neurolink",
-  "version": "12.44.1",
+  "version": "12.44.3",
   "packageManager": "pnpm@10.15.1",
   "description": "The pipe layer of an AI nervous system: one interface connecting provider neurons to your application, across three inference types — generate, stream and decide. `decide` returns typed, calibrated judgements from a non-generative model (~400ms, ~$0.00002/call) for routing, tool selection and context budgeting. MCP-native (4 transports), voice TTS/STT/realtime, RAG, agents, memory, compaction, 9 observability exporters. OpenAI · Anthropic · Gemini · Bedrock · Azure · Ollama · TypeSafe Jev and more.",
   "author": {