feat(providers): integrate DeepSeek, NVIDIA NIM, LM Studio, llama.cpp

Add four OpenAI-compatible providers via the AI SDK's createOpenAI().chat()
endpoint, each with full BaseProvider semantics, OTEL tracing, error
formatting, and validateConfiguration support.

Providers:
- DeepSeek (api.deepseek.com): cloud, requires DEEPSEEK_API_KEY.
- NVIDIA NIM (integrate.api.nvidia.com): cloud, requires NVIDIA_NIM_API_KEY,
  with NIM-specific extras (top_k, min_p, repetition_penalty, reasoning_budget,
  chat_template) passed via providerOptions.openai.body and a retry-on-400
  path that strips unsupported fields.
- LM Studio (localhost:1234): local, auto-discovers loaded model from
  /v1/models. Optional LM_STUDIO_API_KEY for users behind an auth proxy.
- llama.cpp (localhost:8080): local, auto-discovers loaded model. /health
  validation has 3x retry. Optional LLAMACPP_API_KEY for proxied setups.

Credential pass-through verified end-to-end: per-call options.credentials >
NeuroLink instance credentials > env vars > documented defaults.

Pricing changes (src/lib/utils/pricing.ts):
- Add 4 provider entries with symbolic local rates ($1/M tokens) so cost
  attribution reports a non-zero value after the 6-decimal rounding step.
- Treat the `_default` map key as a provider-level fallback (filtered from
  prefix matches, used as last resort) so providers that don't enumerate
  per-model pricing still get a non-undefined rate.

Context-window root-cause fix (src/lib/constants/contextWindows.ts):
- Clamp getOutputReserve to 80% of the context window. Previously, callers
  that passed maxTokens === contextWindow (or larger) got a 0-token input
  budget and every request failed with "Budget: 0 tokens" before being sent.
  Affected 12 test suites and any user passing an oversized maxTokens.

BaseProvider change (src/lib/core/baseProvider.ts):
- Drop `readonly` on `modelName` so providers that auto-discover the model
  via /v1/models (lm-studio, llamacpp) can update it after construction,
  ensuring TelemetryHandler/MessageBuilder cache the resolved name and
  result.model is never empty.

Test infrastructure fixes:
- Twelve test files (memory, context, evaluation, mcp, mcp-http, ppt,
  observability, workflow, tts, media-gen, session-memory-bugs,
  evaluation-scoring) had a shared `PROVIDER_MAX_TOKENS[provider] || 8192`
  default. For local providers with 8K context this set maxTokens to the
  full window → 0 input budget. Fallback lowered to 1024 and the four new
  providers added explicitly to the per-suite map.
- continuous-test-suite-providers.ts: rename `Gemini 3 - DisableTools` test
  to `DisableTools` and make it provider-agnostic (uses TEST_PROVIDER instead
  of hard-coded vertex+gemini-3-flash-preview). The Observability Spans test
  now SKIPs gracefully for Pipeline-A providers (AI SDK + Langfuse OTEL),
  which intentionally do not emit Pipeline-B spans.
- continuous-test-suite-context.ts: Context Compaction Vertex Flash/Pro
  tests now respect TEST_PROVIDER with vertex as fallback.
- continuous-test-suite-evaluation.ts: dimension-specific judge framing in
  scoreAnswerOnDimension so RAGAS Context Precision actually evaluates the
  context (not the answer).
- continuous-test-suite-client.ts: pass timeout: TEST_CONFIG.timeout to
  createServer (was using the undocumented 30s default).
- continuous-test-suite-observability.ts: read TEST_TIMEOUT_MS from env so
  CLI-subprocess tests can be tuned for slow local providers.

Tooling:
- New test/continuous-test-suite-new-providers.ts (10 sections × 4 providers).
- New test/run-provider-matrix.sh runner (bash 3.2 compatible).
- Add test:dynamic, test:proxy, test:bugfixes, test:new-providers scripts.
- Add js-yaml + @types/js-yaml devDeps (proxy Config Loading test fixture).

Documentation:
- docs/provider-integration/ — 14 markdown files covering architecture,
  per-provider integration notes, testing methodology, the failure
  investigation trail, and the PR analysis itself.

Verification:
- `pnpm run check` 0 errors / 0 warnings (3632 files)
- `pnpm run lint` 0 errors / 18 warnings (all pre-existing in unrelated files)
- `pnpm run build` clean
- All 4 providers smoke-tested end-to-end via env, instance, and per-call
  credential paths.

Author: murdore

.env.example

@@ -805,3 +805,62 @@ PICOVOICE_ACCESS_KEY=
 # Optional: override LLM provider and model used by the voice pipeline
 VOICE_LLM_MODEL=gpt-4o-automatic
 VOICE_LLM_PROVIDER=azure
+
+# =============================================================================
+# DEEPSEEK CONFIGURATION
+# =============================================================================
+# Get an API key at https://platform.deepseek.com/api_keys
+DEEPSEEK_API_KEY=
+# Optional: override default model (deepseek-chat | deepseek-reasoner)
+DEEPSEEK_MODEL=deepseek-chat
+# Optional: override default base URL
+# DEEPSEEK_BASE_URL=https://api.deepseek.com
+
+# =============================================================================
+# NVIDIA NIM CONFIGURATION
+# =============================================================================
+# Get an API key at https://build.nvidia.com/settings/api-keys
+NVIDIA_NIM_API_KEY=
+# Optional: override default model (browse https://build.nvidia.com/models)
+NVIDIA_NIM_MODEL=meta/llama-3.3-70b-instruct
+# Optional: override default base URL (use for self-hosted NIM)
+# NVIDIA_NIM_BASE_URL=https://integrate.api.nvidia.com/v1
+# Optional NIM extras (rarely needed, leave commented)
+# NVIDIA_NIM_TOP_K=
+# NVIDIA_NIM_MIN_P=
+# NVIDIA_NIM_REPETITION_PENALTY=
+# NVIDIA_NIM_MIN_TOKENS=
+# NVIDIA_NIM_CHAT_TEMPLATE=
+
+# =============================================================================
+# LM STUDIO CONFIGURATION (local provider)
+# =============================================================================
+# Install LM Studio: https://lmstudio.ai/
+# Load a model in the app and click "Start Server"
+LM_STUDIO_BASE_URL=http://localhost:1234/v1
+# Optional: explicit model id (blank = auto-discover from /v1/models)
+LM_STUDIO_MODEL=
+# Optional: API key (not required for stock LM Studio; use only when running
+# behind an auth-proxying reverse-proxy)
+# LM_STUDIO_API_KEY=
+
+# =============================================================================
+# LLAMA.CPP CONFIGURATION (local provider)
+# =============================================================================
+# Run: ./llama-server -m model.gguf --port 8080 (add --jinja for tool support)
+LLAMACPP_BASE_URL=http://localhost:8080/v1
+# Optional: explicit model id (blank = use whatever model llama-server has loaded)
+LLAMACPP_MODEL=
+# Optional: API key (not required for stock llama-server; use only when running
+# behind an auth-proxying reverse-proxy)
+# LLAMACPP_API_KEY=
+
+# =============================================================================
+# TEST-ONLY CREDENTIALS (used by test/continuous-test-suite-credentials.ts and
+# test/continuous-test-suite-new-providers.ts to verify per-call overrides
+# without depending on the runtime env vars above)
+# =============================================================================
+# TEST_DEEPSEEK_API_KEY=
+# TEST_NVIDIA_NIM_API_KEY=
+# TEST_LM_STUDIO_BASE_URL=
+# TEST_LLAMACPP_BASE_URL=

README.md

@@ -26,7 +26,7 @@ for await (const chunk of result.stream) {
 
 ## 🧠 What is NeuroLink?
 
-**NeuroLink is the universal AI integration platform that unifies 13 major AI providers and 100+ models under one consistent API.**
+**NeuroLink is the universal AI integration platform that unifies 17 major AI providers and 100+ models under one consistent API.**
 
 Extracted from production systems at Juspay and battle-tested at enterprise scale, NeuroLink provides a production-ready solution for integrating AI into any application. Whether you're building with OpenAI, Anthropic, Google, AWS Bedrock, Azure, or any of our 13 supported providers, NeuroLink gives you a single, consistent interface that works everywhere.
 
@@ -195,7 +195,7 @@ npx @juspay/neurolink --help
 
 ### Configuration
 
-NeuroLink works with 13+ AI providers. You'll need at least one API key to get started:
+NeuroLink works with 17+ AI providers. You'll need at least one API key to get started:
 
 **Option 1: Interactive Setup (Recommended)**
 
@@ -382,7 +382,7 @@ const result = await neurolink.generate({
 ### Next Steps
 
 - **[Complete Documentation](https://docs.neurolink.ink)** - Comprehensive guides and API reference
-- **[Provider Setup Guide](docs/getting-started/provider-setup.md)** - Configure all 13 providers
+- **[Provider Setup Guide](docs/getting-started/provider-setup.md)** - Configure all 17 providers
 - **[SDK API Reference](docs/sdk/api-reference.md)** - Full TypeScript API documentation
 - **[CLI Command Reference](docs/cli/commands.md)** - Complete CLI documentation
 - **[Example Projects](docs/examples/index.md)** - Real-world integration examples
@@ -414,7 +414,7 @@ NeuroLink is a comprehensive AI development platform. Every feature below is pro
 
 ### 🤖 AI Provider Integration
 
-**13 providers unified under one API** - Switch providers with a single parameter change.
+**17 providers unified under one API** - Switch providers with a single parameter change.
 
 | Provider              | Models                                             | Free Tier       | Tool Support | Status        | Documentation                                                                                                                 |
 | --------------------- | -------------------------------------------------- | --------------- | ------------ | ------------- | ----------------------------------------------------------------------------------------------------------------------------- |
@@ -431,6 +431,10 @@ NeuroLink is a comprehensive AI development platform. Every feature below is pro
 | **Ollama**            | Local models (Llama, Mistral)                      | ✅ Free (Local) | ⚠️ Partial   | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#ollama)                                                                  |
 | **OpenAI Compatible** | Any OpenAI-compatible endpoint                     | Varies          | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#openai-compatible)                                                       |
 | **OpenRouter**        | 200+ Models via OpenRouter                         | Varies          | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/providers/openrouter.md)                                                                   |
+| **DeepSeek**          | deepseek-chat (V3), deepseek-reasoner (R1)         | ❌              | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#deepseek)                                                                |
+| **NVIDIA NIM**        | Llama 3.3 70B, 400+ catalog models                 | ❌              | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#nvidia-nim)                                                              |
+| **LM Studio**         | Any model loaded in LM Studio (local)              | ✅ Free (Local) | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#lm-studio)                                                               |
+| **llama.cpp**         | Any GGUF model served by llama-server (local)      | ✅ Free (Local) | ✅ Full      | ✅ Production | [Setup Guide](docs/getting-started/provider-setup.md#llamacpp)                                                                |
 
 **[📖 Provider Comparison Guide](docs/reference/provider-comparison.md)** - Detailed feature matrix and selection criteria
 **[🔬 Provider Feature Compatibility](docs/reference/provider-feature-compatibility.md)** - Test-based compatibility reference for all 19 features across 13 providers
@@ -589,7 +593,7 @@ const result = await neurolink.generate({
 - **ProcessorRegistry** - Priority-based processor selection with fallback
 - **OWASP Security** - HTML/SVG sanitization prevents XSS attacks
 - **Auto-detection** - FileDetector identifies file types by extension and content
-- **Provider-agnostic** - All processors work across all 13 AI providers
+- **Provider-agnostic** - All processors work across all 17 AI providers
 
 **[📖 File Processors Guide](docs/features/file-processors.md)** - Complete reference for all file types
 
@@ -731,7 +735,7 @@ Run AI-powered workflows directly in GitHub Actions with 13 provider support and
 
 | Feature                | Description                                                                               |
 | ---------------------- | ----------------------------------------------------------------------------------------- |
-| **Multi-Provider**     | 13 providers with unified interface                                                       |
+| **Multi-Provider**     | 17 providers with unified interface                                                       |
 | **PR/Issue Comments**  | Auto-post AI responses with intelligent updates                                           |
 | **Multimodal Support** | Attach images, PDFs, CSVs, Excel, Word, JSON, YAML, XML, HTML, SVG, code files to prompts |
 | **Cost Tracking**      | Built-in analytics and quality evaluation                                                 |
@@ -924,7 +928,7 @@ Full command and API breakdown lives in [`docs/cli/commands.md`](docs/cli/comman
 
 | Capability               | Highlights                                                                                                               |
 | ------------------------ | ------------------------------------------------------------------------------------------------------------------------ |
-| **Provider unification** | 13+ providers with automatic fallback, cost-aware routing, provider orchestration (Q3).                                  |
+| **Provider unification** | 17+ providers with automatic fallback, cost-aware routing, provider orchestration (Q3).                                  |
 | **Multimodal pipeline**  | Stream images + CSV data + PDF documents across providers with local/remote assets. Auto-detection for mixed file types. |
 | **Quality & governance** | Auto-evaluation engine (Q3), guardrails middleware (Q4), HITL workflows (Q4), audit logging.                             |
 | **Memory & context**     | Conversation memory, Redis history export (Q4), context summarization (Q4).                                              |

docs-site/scripts/sync-docs.ts

@@ -1035,6 +1035,14 @@ const LINK_MAPPINGS: Record<string, string> = {
   "providers/openrouter": "/getting-started/providers/openrouter",
   "providers/anthropic": "/getting-started/providers/anthropic",
   anthropic: "/getting-started/providers/anthropic",
+  // Bare provider names so sibling-relative `./<name>.md` links inside
+  // docs/getting-started/providers/ resolve to the canonical Docusaurus path
+  // instead of falling through to the generic `/docs/<name>` (which 404s).
+  ollama: "/getting-started/providers/ollama",
+  deepseek: "/getting-started/providers/deepseek",
+  "nvidia-nim": "/getting-started/providers/nvidia-nim",
+  "lm-studio": "/getting-started/providers/lm-studio",
+  llamacpp: "/getting-started/providers/llamacpp",
 
   // MCP additional
   "mcp/server-catalog": "/guides/mcp/server-catalog",

docs/getting-started/environment-variables.md

@@ -861,12 +861,138 @@ npx @juspay/neurolink sagemaker setup
 
 ---
 
+### 12. DeepSeek
+
+#### Required Variables
+
+```bash
+DEEPSEEK_API_KEY="sk-your-deepseek-api-key"
+```
+
+#### Optional Variables
+
+```bash
+DEEPSEEK_MODEL="deepseek-chat"                    # Default: deepseek-chat (use deepseek-reasoner for R1)
+DEEPSEEK_BASE_URL="https://api.deepseek.com"      # Default: DeepSeek API
+```
+
+#### How to Get DeepSeek API Key
+
+1. Visit [DeepSeek Platform](https://platform.deepseek.com/api_keys)
+2. Sign up or log in to your account
+3. Navigate to **API Keys** section
+4. Click **Create API Key**
+5. Copy the key
+
+#### Supported Models
+
+- `deepseek-chat` (default) - DeepSeek V3, high-quality general chat
+- `deepseek-reasoner` - DeepSeek R1, extended chain-of-thought reasoning
+
+---
+
+### 13. NVIDIA NIM
+
+#### Required Variables
+
+```bash
+NVIDIA_NIM_API_KEY="nvapi-your-nvidia-api-key"
+```
+
+#### Optional Variables
+
+```bash
+NVIDIA_NIM_MODEL="meta/llama-3.3-70b-instruct"              # Default model
+NVIDIA_NIM_BASE_URL="https://integrate.api.nvidia.com/v1"   # Default: NVIDIA cloud API (override for self-hosted NIM)
+```
+
+#### NIM-Specific Extras (rarely needed)
+
+```bash
+# Sampling extras passed as request body extensions
+NVIDIA_NIM_TOP_K=                      # Integer, -1 = disabled (default)
+NVIDIA_NIM_MIN_P=                      # Float, 0 = disabled (default)
+NVIDIA_NIM_REPETITION_PENALTY=         # Float, 1.0 = disabled (default)
+NVIDIA_NIM_MIN_TOKENS=                 # Integer, 0 = disabled (default)
+NVIDIA_NIM_CHAT_TEMPLATE=              # Override model chat template string (advanced)
+```
+
+#### How to Get NVIDIA NIM API Key
+
+1. Visit [NVIDIA Build](https://build.nvidia.com/)
+2. Sign in with your NVIDIA developer account
+3. Open **Settings → API Keys**
+4. Generate a new API key (Bearer token)
+
+#### Supported Models
+
+- `meta/llama-3.3-70b-instruct` (default) - Llama 3.3 70B Instruct
+- Any model listed at [build.nvidia.com/models](https://build.nvidia.com/models)
+
+---
+
+### 14. LM Studio (Local)
+
+LM Studio is a local provider — no API key is required for standard installations.
+
+#### Optional Variables
+
+```bash
+LM_STUDIO_BASE_URL="http://localhost:1234/v1"    # Default: local LM Studio server
+LM_STUDIO_MODEL=""                               # Blank = auto-discover from /v1/models
+# LM_STUDIO_API_KEY=                             # Only set when running behind an auth-proxying reverse-proxy
+```
+
+#### How to Set Up LM Studio
+
+1. Install LM Studio from [lmstudio.ai](https://lmstudio.ai/)
+2. Open LM Studio and download a model (e.g., Llama 3.2 3B Instruct)
+3. Click **Local Server** → **Start Server**
+4. The server starts at `http://localhost:1234/v1` by default
+5. NeuroLink auto-discovers the loaded model; no `LM_STUDIO_MODEL` needed
+
+**Note:** `LM_STUDIO_API_KEY` is only needed if you run LM Studio behind an authenticating reverse-proxy. Vanilla local installs do not require an API key.
+
+---
+
+### 15. llama.cpp (Local)
+
+llama.cpp (llama-server) is a local provider — no API key is required for standard installations.
+
+#### Optional Variables
+
+```bash
+LLAMACPP_BASE_URL="http://localhost:8080/v1"     # Default: local llama-server
+LLAMACPP_MODEL=""                                # Blank = use whatever model llama-server has loaded
+# LLAMACPP_API_KEY=                              # Only set when running behind an auth-proxying reverse-proxy
+```
+
+#### How to Set Up llama.cpp
+
+1. Build llama.cpp from source: [github.com/ggerganov/llama.cpp](https://github.com/ggerganov/llama.cpp#build)
+2. Download a GGUF model file
+3. Start llama-server:
+
+   ```bash
+   # Basic usage
+   ./llama-server -m model.gguf --port 8080
+
+   # With tool/function-call support (required for MCP tools)
+   ./llama-server -m model.gguf --port 8080 --jinja
+   ```
+
+4. NeuroLink auto-discovers the loaded model; no `LLAMACPP_MODEL` needed
+
+**Note:** `LLAMACPP_API_KEY` is only needed if you run llama-server behind an authenticating reverse-proxy. Vanilla local installs do not require an API key.
+
+---
+
 ## 🔧 Configuration Examples
 
 ### Complete .env File Example
 
 ```bash
-# NeuroLink Environment Configuration - All 11 Providers
+# NeuroLink Environment Configuration - All 15 Providers
 
 # OpenAI Configuration
 OPENAI_API_KEY="sk-proj-your-openai-key"
@@ -926,6 +1052,26 @@ MISTRAL_MODEL="mistral-small"
 LITELLM_BASE_URL="http://localhost:4000"
 LITELLM_API_KEY="sk-anything"
 LITELLM_MODEL="openai/gpt-4o-mini"
+
+# DeepSeek Configuration
+DEEPSEEK_API_KEY="sk-your-deepseek-key"
+DEEPSEEK_MODEL="deepseek-chat"
+# DEEPSEEK_BASE_URL=https://api.deepseek.com
+
+# NVIDIA NIM Configuration
+NVIDIA_NIM_API_KEY="nvapi-your-nvidia-key"
+NVIDIA_NIM_MODEL="meta/llama-3.3-70b-instruct"
+# NVIDIA_NIM_BASE_URL=https://integrate.api.nvidia.com/v1
+
+# LM Studio Configuration (local — no API key required)
+LM_STUDIO_BASE_URL="http://localhost:1234/v1"
+# LM_STUDIO_MODEL=                 # blank = auto-discover
+# LM_STUDIO_API_KEY=               # only for reverse-proxy setups
+
+# llama.cpp Configuration (local — no API key required)
+LLAMACPP_BASE_URL="http://localhost:8080/v1"
+# LLAMACPP_MODEL=                  # blank = auto-discover
+# LLAMACPP_API_KEY=                # only for reverse-proxy setups
 ```
 
 ### Docker/Container Configuration