feat: add NEAR AI Cloud provider

.env.example

@@ -23,6 +23,9 @@
 # OpenRouter (multi-provider gateway)
 # OPENROUTER_API_KEY=sk-or-...
 
+# NEAR AI Cloud (TEE inference)
+# NEARAI_API_KEY=...
+
 # Together AI
 # TOGETHER_API_KEY=...
 

README.md

@@ -357,11 +357,11 @@ For production workloads, use the [WhatsApp Cloud API](https://developers.facebo
 
 ---
 
-## 27 LLM Providers, 123+ Models
+## LLM Providers, 128+ Models
 
-3 native drivers (Anthropic, Gemini, OpenAI-compatible) route to 27 providers:
+3 native drivers (Anthropic, Gemini, OpenAI-compatible) route to providers including:
 
-Anthropic, Gemini, OpenAI, Groq, DeepSeek, OpenRouter, Together, Mistral, Fireworks, Cohere, Perplexity, xAI, AI21, Cerebras, SambaNova, HuggingFace, Replicate, Ollama, vLLM, LM Studio, Qwen, MiniMax, Zhipu, Moonshot, Qianfan, Bedrock, and more.
+Anthropic, Gemini, OpenAI, Groq, DeepSeek, OpenRouter, NEAR AI Cloud, Together, Mistral, Fireworks, Cohere, Perplexity, xAI, AI21, Cerebras, SambaNova, HuggingFace, Replicate, Ollama, vLLM, LM Studio, Qwen, MiniMax, Zhipu, Moonshot, Qianfan, Bedrock, and more.
 
 Intelligent routing with task complexity scoring, automatic fallback, cost tracking, and per-model pricing.
 

crates/openfang-runtime/src/drivers/mod.rs

@@ -2,7 +2,7 @@
 //!
 //! Contains drivers for Anthropic Claude, Google Gemini, OpenAI-compatible APIs, and more.
 //! Supports: Anthropic, Gemini, OpenAI, Groq, OpenRouter, DeepSeek, Together,
-//! Mistral, Fireworks, Ollama, vLLM, Chutes.ai, and any OpenAI-compatible endpoint.
+//! Mistral, Fireworks, NEAR AI Cloud, Ollama, vLLM, Chutes.ai, and any OpenAI-compatible endpoint.
 
 pub mod anthropic;
 pub mod bedrock;
@@ -19,11 +19,12 @@ use openfang_types::model_catalog::{
     AI21_BASE_URL, ANTHROPIC_BASE_URL, AZURE_OPENAI_BASE_URL, CEREBRAS_BASE_URL, CHUTES_BASE_URL,
     COHERE_BASE_URL, DEEPSEEK_BASE_URL, FIREWORKS_BASE_URL, GEMINI_BASE_URL, GROQ_BASE_URL,
     HUGGINGFACE_BASE_URL, KIMI_CODING_BASE_URL, LEMONADE_BASE_URL, LMSTUDIO_BASE_URL,
-    MINIMAX_BASE_URL, MISTRAL_BASE_URL, MOONSHOT_BASE_URL, NOVITA_BASE_URL, NVIDIA_NIM_BASE_URL,
-    OLLAMA_BASE_URL, OPENAI_BASE_URL, OPENROUTER_BASE_URL, PERPLEXITY_BASE_URL, QIANFAN_BASE_URL,
-    QWEN_BASE_URL, REPLICATE_BASE_URL, REQUESTY_BASE_URL, SAMBANOVA_BASE_URL, TOGETHER_BASE_URL,
-    VENICE_BASE_URL, VLLM_BASE_URL, VOLCENGINE_BASE_URL, VOLCENGINE_CODING_BASE_URL, XAI_BASE_URL,
-    ZAI_BASE_URL, ZAI_CODING_BASE_URL, ZHIPU_BASE_URL, ZHIPU_CODING_BASE_URL,
+    MINIMAX_BASE_URL, MISTRAL_BASE_URL, MOONSHOT_BASE_URL, NEARAI_BASE_URL, NOVITA_BASE_URL,
+    NVIDIA_NIM_BASE_URL, OLLAMA_BASE_URL, OPENAI_BASE_URL, OPENROUTER_BASE_URL,
+    PERPLEXITY_BASE_URL, QIANFAN_BASE_URL, QWEN_BASE_URL, REPLICATE_BASE_URL, REQUESTY_BASE_URL,
+    SAMBANOVA_BASE_URL, TOGETHER_BASE_URL, VENICE_BASE_URL, VLLM_BASE_URL, VOLCENGINE_BASE_URL,
+    VOLCENGINE_CODING_BASE_URL, XAI_BASE_URL, ZAI_BASE_URL, ZAI_CODING_BASE_URL, ZHIPU_BASE_URL,
+    ZHIPU_CODING_BASE_URL,
 };
 use std::sync::Arc;
 
@@ -276,6 +277,11 @@ fn provider_defaults(provider: &str) -> Option<ProviderDefaults> {
             api_key_env: "CHUTES_API_KEY",
             key_required: true,
         }),
+        "nearai" | "near-ai" => Some(ProviderDefaults {
+            base_url: NEARAI_BASE_URL,
+            api_key_env: "NEARAI_API_KEY",
+            key_required: true,
+        }),
         "venice" => Some(ProviderDefaults {
             base_url: VENICE_BASE_URL,
             api_key_env: "VENICE_API_KEY",
@@ -329,6 +335,7 @@ fn provider_defaults(provider: &str) -> Option<ProviderDefaults> {
 /// - `xai` — xAI (Grok)
 /// - `replicate` — Replicate
 /// - `chutes` — Chutes.ai (serverless open-source model inference)
+/// - `nearai` — NEAR AI Cloud TEE inference
 /// - Any custom provider with `base_url` set uses OpenAI-compatible format
 pub fn create_driver(config: &DriverConfig) -> Result<Arc<dyn LlmDriver>, LlmError> {
     let provider = config.provider.as_str();
@@ -589,7 +596,7 @@ pub fn create_driver(config: &DriverConfig) -> Result<Arc<dyn LlmDriver>, LlmErr
             "Unknown provider '{}'. Supported: anthropic, gemini, openai, azure, bedrock, groq, \
              openrouter, deepseek, together, mistral, fireworks, ollama, vllm, lmstudio, \
              perplexity, cohere, ai21, cerebras, sambanova, huggingface, xai, replicate, \
-             github-copilot, chutes, venice, nvidia, codex, claude-code. \
+             github-copilot, chutes, nearai, venice, nvidia, codex, claude-code. \
              Or set base_url for a custom OpenAI-compatible endpoint.",
             provider
         ),
@@ -631,6 +638,7 @@ pub fn detect_available_provider() -> Option<(&'static str, &'static str, &'stat
             "PERPLEXITY_API_KEY",
         ),
         ("cohere", "command-r-plus", "COHERE_API_KEY"),
+        ("nearai", "zai-org/GLM-5.1-FP8", "NEARAI_API_KEY"),
         ("novita", "moonshotai/kimi-k2.5", "NOVITA_API_KEY"),
     ];
     for &(provider, model, env_var) in PROBE_ORDER {
@@ -687,6 +695,7 @@ pub fn known_providers() -> &'static [&'static str] {
         "qianfan",
         "volcengine",
         "chutes",
+        "nearai",
         "venice",
         "nvidia",
         "novita",
@@ -839,13 +848,14 @@ mod tests {
         assert!(providers.contains(&"qianfan"));
         assert!(providers.contains(&"volcengine"));
         assert!(providers.contains(&"chutes"));
+        assert!(providers.contains(&"nearai"));
         assert!(providers.contains(&"nvidia"));
         assert!(providers.contains(&"novita"));
         assert!(providers.contains(&"codex"));
         assert!(providers.contains(&"claude-code"));
         assert!(providers.contains(&"qwen-code"));
         assert!(providers.contains(&"azure"));
-        assert_eq!(providers.len(), 38);
+        assert_eq!(providers.len(), 39);
     }
 
     #[test]
@@ -902,6 +912,56 @@ mod tests {
         assert!(d.key_required);
     }
 
+    #[test]
+    fn test_provider_defaults_nearai() {
+        let d = provider_defaults("nearai").unwrap();
+        assert_eq!(d.base_url, "https://cloud-api.near.ai/v1");
+        assert_eq!(d.api_key_env, "NEARAI_API_KEY");
+        assert!(d.key_required);
+    }
+
+    #[test]
+    fn test_provider_defaults_near_ai_alias() {
+        let d = provider_defaults("near-ai").unwrap();
+        assert_eq!(d.base_url, "https://cloud-api.near.ai/v1");
+        assert_eq!(d.api_key_env, "NEARAI_API_KEY");
+        assert!(d.key_required);
+    }
+
+    #[test]
+    fn test_nearai_provider_with_env_key() {
+        let _env_lock = ENV_LOCK.lock().unwrap();
+        let unique_key = "test-nearai-key-12345";
+        let _env = EnvVarGuard::set("NEARAI_API_KEY", unique_key);
+        let config = DriverConfig {
+            provider: "nearai".to_string(),
+            api_key: None,
+            base_url: None,
+            skip_permissions: true,
+            subprocess_timeout_secs: None,
+        };
+        let driver = create_driver(&config);
+        assert!(
+            driver.is_ok(),
+            "NEAR AI provider with env var should succeed"
+        );
+    }
+
+    #[test]
+    fn test_nearai_provider_no_key_errors() {
+        let _env_lock = ENV_LOCK.lock().unwrap();
+        let _env = EnvVarGuard::remove("NEARAI_API_KEY");
+        let config = DriverConfig {
+            provider: "nearai".to_string(),
+            api_key: None,
+            base_url: None,
+            skip_permissions: true,
+            subprocess_timeout_secs: None,
+        };
+        let driver = create_driver(&config);
+        assert!(driver.is_err());
+    }
+
     #[test]
     fn test_novita_provider_with_env_key() {
         let _env_lock = ENV_LOCK.lock().unwrap();

crates/openfang-runtime/src/model_catalog.rs

@@ -1,18 +1,18 @@
 //! Model catalog — registry of known models with metadata, pricing, and auth detection.
 //!
-//! Provides a comprehensive catalog of 130+ builtin models across 28 providers,
+//! Provides a comprehensive catalog of 130+ builtin models across 43 providers,
 //! with alias resolution, auth status detection, and pricing lookups.
 
 use openfang_types::model_catalog::{
     AuthStatus, ModelCatalogEntry, ModelTier, ProviderInfo, AI21_BASE_URL, ANTHROPIC_BASE_URL,
     AZURE_OPENAI_BASE_URL, BEDROCK_BASE_URL, CEREBRAS_BASE_URL, CHUTES_BASE_URL, COHERE_BASE_URL,
     DEEPSEEK_BASE_URL, FIREWORKS_BASE_URL, GEMINI_BASE_URL, GITHUB_COPILOT_BASE_URL, GROQ_BASE_URL,
     HUGGINGFACE_BASE_URL, KIMI_CODING_BASE_URL, LEMONADE_BASE_URL, LMSTUDIO_BASE_URL,
-    MINIMAX_BASE_URL, MISTRAL_BASE_URL, MOONSHOT_BASE_URL, NVIDIA_NIM_BASE_URL, OLLAMA_BASE_URL,
-    OPENAI_BASE_URL, OPENROUTER_BASE_URL, PERPLEXITY_BASE_URL, QIANFAN_BASE_URL, QWEN_BASE_URL,
-    REPLICATE_BASE_URL, REQUESTY_BASE_URL, SAMBANOVA_BASE_URL, TOGETHER_BASE_URL, VENICE_BASE_URL,
-    VLLM_BASE_URL, VOLCENGINE_BASE_URL, VOLCENGINE_CODING_BASE_URL, XAI_BASE_URL, ZAI_BASE_URL,
-    ZAI_CODING_BASE_URL, ZHIPU_BASE_URL, ZHIPU_CODING_BASE_URL,
+    MINIMAX_BASE_URL, MISTRAL_BASE_URL, MOONSHOT_BASE_URL, NEARAI_BASE_URL, NVIDIA_NIM_BASE_URL,
+    OLLAMA_BASE_URL, OPENAI_BASE_URL, OPENROUTER_BASE_URL, PERPLEXITY_BASE_URL, QIANFAN_BASE_URL,
+    QWEN_BASE_URL, REPLICATE_BASE_URL, REQUESTY_BASE_URL, SAMBANOVA_BASE_URL, TOGETHER_BASE_URL,
+    VENICE_BASE_URL, VLLM_BASE_URL, VOLCENGINE_BASE_URL, VOLCENGINE_CODING_BASE_URL, XAI_BASE_URL,
+    ZAI_BASE_URL, ZAI_CODING_BASE_URL, ZHIPU_BASE_URL, ZHIPU_CODING_BASE_URL,
 };
 use std::collections::HashMap;
 
@@ -790,6 +790,16 @@ fn builtin_providers() -> Vec<ProviderInfo> {
             auth_status: AuthStatus::Missing,
             model_count: 0,
         },
+        // ── NEAR AI Cloud ───────────────────────────────────────────
+        ProviderInfo {
+            id: "nearai".into(),
+            display_name: "NEAR AI Cloud".into(),
+            api_key_env: "NEARAI_API_KEY".into(),
+            base_url: NEARAI_BASE_URL.into(),
+            key_required: true,
+            auth_status: AuthStatus::Missing,
+            model_count: 0,
+        },
         // ── Venice.ai ────────────────────────────────────────────────
         ProviderInfo {
             id: "venice".into(),
@@ -1030,6 +1040,9 @@ fn builtin_aliases() -> HashMap<String, String> {
         ("codex-o4", "codex/o4-mini"),
         // NVIDIA NIM aliases
         ("nemotron", "nvidia/llama-3.1-nemotron-70b-instruct"),
+        // NEAR AI Cloud aliases
+        ("nearai", "zai-org/GLM-5.1-FP8"),
+        ("nearai-glm", "zai-org/GLM-5.1-FP8"),
         // Venice aliases
         ("venice", "venice-uncensored"),
         // Claude Code aliases
@@ -4023,6 +4036,83 @@ fn builtin_models() -> Vec<ModelCatalogEntry> {
             aliases: vec!["chutes-llama-70b".into()],
         },
         // ══════════════════════════════════════════════════════════════
+        // NEAR AI Cloud TEE inference (5)
+        // ══════════════════════════════════════════════════════════════
+        ModelCatalogEntry {
+            id: "zai-org/GLM-5.1-FP8".into(),
+            display_name: "GLM 5.1 (NEAR AI TEE)".into(),
+            provider: "nearai".into(),
+            tier: ModelTier::Smart,
+            context_window: 202_752,
+            max_output_tokens: 8_192,
+            input_cost_per_m: 0.85,
+            output_cost_per_m: 3.3,
+            supports_tools: true,
+            supports_vision: false,
+            supports_streaming: true,
+            aliases: vec![
+                "nearai".into(),
+                "nearai-glm".into(),
+                "nearai-glm-5.1".into(),
+            ],
+        },
+        ModelCatalogEntry {
+            id: "Qwen/Qwen3.5-122B-A10B".into(),
+            display_name: "Qwen3.5 122B A10B (NEAR AI TEE)".into(),
+            provider: "nearai".into(),
+            tier: ModelTier::Smart,
+            context_window: 131_072,
+            max_output_tokens: 8_192,
+            input_cost_per_m: 0.4,
+            output_cost_per_m: 3.2,
+            supports_tools: true,
+            supports_vision: false,
+            supports_streaming: true,
+            aliases: vec!["nearai-qwen3.5".into(), "nearai-qwen3.5-122b".into()],
+        },
+        ModelCatalogEntry {
+            id: "Qwen/Qwen3.6-35B-A3B-FP8".into(),
+            display_name: "Qwen 3.6 35B A3B FP8 (NEAR AI TEE)".into(),
+            provider: "nearai".into(),
+            tier: ModelTier::Balanced,
+            context_window: 262_144,
+            max_output_tokens: 8_192,
+            input_cost_per_m: 0.17,
+            output_cost_per_m: 1.1,
+            supports_tools: true,
+            supports_vision: false,
+            supports_streaming: true,
+            aliases: vec!["nearai-qwen3.6".into(), "nearai-qwen3.6-35b".into()],
+        },
+        ModelCatalogEntry {
+            id: "Qwen/Qwen3-VL-30B-A3B-Instruct".into(),
+            display_name: "Qwen3 VL 30B A3B Instruct (NEAR AI TEE)".into(),
+            provider: "nearai".into(),
+            tier: ModelTier::Balanced,
+            context_window: 256_000,
+            max_output_tokens: 8_192,
+            input_cost_per_m: 0.15,
+            output_cost_per_m: 0.55,
+            supports_tools: true,
+            supports_vision: true,
+            supports_streaming: true,
+            aliases: vec!["nearai-qwen-vl".into(), "nearai-qwen3-vl".into()],
+        },
+        ModelCatalogEntry {
+            id: "google/gemma-4-31B-it".into(),
+            display_name: "Gemma 4 31B Instruct (NEAR AI TEE)".into(),
+            provider: "nearai".into(),
+            tier: ModelTier::Fast,
+            context_window: 262_144,
+            max_output_tokens: 8_192,
+            input_cost_per_m: 0.13,
+            output_cost_per_m: 0.4,
+            supports_tools: true,
+            supports_vision: false,
+            supports_streaming: true,
+            aliases: vec!["nearai-gemma".into(), "nearai-gemma-4".into()],
+        },
+        // ══════════════════════════════════════════════════════════════
         // Venice.ai (3)
         // ══════════════════════════════════════════════════════════════
         ModelCatalogEntry {
@@ -4083,7 +4173,7 @@ mod tests {
     #[test]
     fn test_catalog_has_providers() {
         let catalog = ModelCatalog::new();
-        assert_eq!(catalog.list_providers().len(), 42);
+        assert_eq!(catalog.list_providers().len(), 43);
     }
 
     #[test]
@@ -4231,6 +4321,34 @@ mod tests {
         assert!(catalog.get_provider("huggingface").is_some());
         assert!(catalog.get_provider("xai").is_some());
         assert!(catalog.get_provider("replicate").is_some());
+        assert!(catalog.get_provider("nearai").is_some());
+    }
+
+    #[test]
+    fn test_nearai_provider_and_models() {
+        let catalog = ModelCatalog::new();
+        let provider = catalog.get_provider("nearai").unwrap();
+        assert_eq!(provider.display_name, "NEAR AI Cloud");
+        assert_eq!(provider.api_key_env, "NEARAI_API_KEY");
+        assert_eq!(provider.base_url, "https://cloud-api.near.ai/v1");
+        assert!(provider.key_required);
+        assert_eq!(provider.model_count, 5);
+
+        let models = catalog.models_by_provider("nearai");
+        assert!(models.iter().all(|m| m.supports_streaming));
+        assert!(models.iter().any(|m| m.id == "zai-org/GLM-5.1-FP8"));
+        assert!(models
+            .iter()
+            .any(|m| m.id == "Qwen/Qwen3-VL-30B-A3B-Instruct" && m.supports_vision));
+    }
+
+    #[test]
+    fn test_nearai_aliases() {
+        let catalog = ModelCatalog::new();
+        let entry = catalog.find_model("nearai").unwrap();
+        assert_eq!(entry.id, "zai-org/GLM-5.1-FP8");
+        assert_eq!(entry.provider, "nearai");
+        assert!(catalog.find_model("nearai-qwen3.6").is_some());
     }
 
     #[test]

crates/openfang-types/src/model_catalog.rs

@@ -30,6 +30,7 @@ pub const SAMBANOVA_BASE_URL: &str = "https://api.sambanova.ai/v1";
 pub const HUGGINGFACE_BASE_URL: &str = "https://api-inference.huggingface.co/v1";
 pub const XAI_BASE_URL: &str = "https://api.x.ai/v1";
 pub const REPLICATE_BASE_URL: &str = "https://api.replicate.com/v1";
+pub const NEARAI_BASE_URL: &str = "https://cloud-api.near.ai/v1";
 pub const VENICE_BASE_URL: &str = "https://api.venice.ai/api/v1";
 pub const NVIDIA_NIM_BASE_URL: &str = "https://integrate.api.nvidia.com/v1";
 

docs/README.md

@@ -1,6 +1,6 @@
 # OpenFang Documentation
 
-Welcome to the OpenFang documentation. OpenFang is the open-source Agent Operating System -- 14 Rust crates, 40 channels, 60 skills, 20 LLM providers, 76 API endpoints, and 16 security systems in a single binary.
+Welcome to the OpenFang documentation. OpenFang is the open-source Agent Operating System -- 14 Rust crates, 40 channels, 60 skills, many LLM providers, 76 API endpoints, and 16 security systems in a single binary.
 
 ---
 
@@ -27,7 +27,7 @@ Welcome to the OpenFang documentation. OpenFang is the open-source Agent Operati
 | Guide | Description |
 |-------|-------------|
 | [Channel Adapters](channel-adapters.md) | 40 messaging channels -- setup, configuration, custom adapters |
-| [LLM Providers](providers.md) | 20 providers, 51 models, 23 aliases -- setup and model routing |
+| [LLM Providers](providers.md) | Provider setup, model catalog, aliases, and routing |
 | [Skills](skill-development.md) | 60 bundled skills, custom skill development, FangHub marketplace |
 | [MCP & A2A](mcp-a2a.md) | Model Context Protocol and Agent-to-Agent protocol integration |
 
@@ -74,9 +74,9 @@ openfang init && openfang start
 | Messaging channels | 40 |
 | Bundled skills | 60 |
 | Built-in tools | 38 |
-| LLM providers | 20 |
-| Models in catalog | 51 |
-| Model aliases | 23 |
+| LLM providers | Many cloud and local providers |
+| Models in catalog | Builtin plus runtime-discovered models |
+| Model aliases | Builtin aliases |
 | API endpoints | 76 |
 | Security systems | 16 |
 | Tests | 967 |
@@ -100,6 +100,7 @@ openfang init && openfang start
 | `GEMINI_API_KEY` | Google Gemini |
 | `GROQ_API_KEY` | Groq (fast Llama/Mixtral) |
 | `DEEPSEEK_API_KEY` | DeepSeek |
+| `NEARAI_API_KEY` | NEAR AI Cloud TEE inference |
 | `XAI_API_KEY` | xAI (Grok) |
 
 Only one provider key is needed to get started. Groq offers a free tier.