feat(config, llm): Add a vllm provider

crates/jp_config/src/model/id.rs

@@ -595,6 +595,9 @@ pub enum ProviderId {
     /// Openrouter provider.
     /// See: <https://openrouter.io>.
     Openrouter,
+    /// vLLM provider: a self-hosted vLLM server with an OpenAI-compatible API.
+    /// See: <https://docs.vllm.ai>.
+    Vllm,
     /// xAI provider.
     /// See: <https://x.ai/api>.
     /// UNIMPLEMENTED.
@@ -619,6 +622,7 @@ impl ProviderId {
             Self::Ollama => "ollama",
             Self::Openai => "openai",
             Self::Openrouter => "openrouter",
+            Self::Vllm => "vllm",
             Self::Xai => "xai",
 
             Self::Test => "test",

crates/jp_config/src/model/id_tests.rs

@@ -42,6 +42,7 @@ fn the_variant_list_leaves_out_the_test_provider() {
         "ollama",
         "openai",
         "openrouter",
+        "vllm",
         "xai"
     ]);
 }

crates/jp_config/src/providers/llm.rs

@@ -8,6 +8,7 @@ pub mod llamacpp;
 pub mod ollama;
 pub mod openai;
 pub mod openrouter;
+pub mod vllm;
 
 use indexmap::IndexMap;
 use schematic::{Config, ConfigError};
@@ -27,6 +28,7 @@ use crate::{
         ollama::{OllamaConfig, PartialOllamaConfig},
         openai::{OpenaiConfig, PartialOpenaiConfig},
         openrouter::{OpenrouterConfig, PartialOpenrouterConfig},
+        vllm::{PartialVllmConfig, VllmConfig},
     },
     util::merge_nested_indexmap,
     validate::Validator,
@@ -83,6 +85,10 @@ pub struct LlmProviderConfig {
     /// Openrouter API configuration.
     #[setting(nested)]
     pub openrouter: OpenrouterConfig,
+
+    /// vLLM API configuration.
+    #[setting(nested)]
+    pub vllm: VllmConfig,
 }
 
 impl Validator for LlmProviderConfig {
@@ -104,6 +110,7 @@ impl AssignKeyValue for PartialLlmProviderConfig {
             _ if kv.p("ollama") => self.ollama.assign(kv)?,
             _ if kv.p("openai") => self.openai.assign(kv)?,
             _ if kv.p("openrouter") => self.openrouter.assign(kv)?,
+            _ if kv.p("vllm") => self.vllm.assign(kv)?,
             _ => return missing_key(&kv),
         }
 
@@ -126,6 +133,7 @@ impl PartialConfigDelta for PartialLlmProviderConfig {
             ollama: self.ollama.delta(next.ollama),
             openai: self.openai.delta(next.openai),
             openrouter: self.openrouter.delta(next.openrouter),
+            vllm: self.vllm.delta(next.vllm),
         }
     }
 
@@ -148,6 +156,7 @@ impl PartialConfigDelta for PartialLlmProviderConfig {
                 &path(prefix, "openrouter"),
                 unsets,
             ),
+            vllm: self.vllm.delta(next.vllm),
         }
     }
 }
@@ -164,6 +173,7 @@ impl FillDefaults for PartialLlmProviderConfig {
             ollama: self.ollama.fill_from(defaults.ollama),
             openai: self.openai.fill_from(defaults.openai),
             openrouter: self.openrouter.fill_from(defaults.openrouter),
+            vllm: self.vllm.fill_from(defaults.vllm),
         }
     }
 }
@@ -184,6 +194,7 @@ impl ToPartial for LlmProviderConfig {
             ollama: self.ollama.to_partial(),
             openai: self.openai.to_partial(),
             openrouter: self.openrouter.to_partial(),
+            vllm: self.vllm.to_partial(),
         }
     }
 }

crates/jp_config/src/providers/llm/vllm.rs

@@ -0,0 +1,77 @@
+//! vLLM API configuration.
+//!
+//! A vLLM server speaks the OpenAI-compatible `/v1/chat/completions` dialect
+//! and checks the Bearer token given to it with `--api-key`.
+//!
+//! ```toml
+//! [providers.llm.vllm]
+//! api_key_env = "VLLM_API_KEY"
+//! base_url = "http://127.0.0.1:8000"
+//! ```
+
+use schematic::Config;
+
+use crate::{
+    assignment::{AssignKeyValue, AssignResult, KvAssignment, missing_key},
+    delta::{PartialConfigDelta, delta_opt},
+    fill::FillDefaults,
+    partial::{ToPartial, partial_opt},
+};
+
+/// vLLM API configuration.
+#[derive(Debug, Clone, PartialEq, Config)]
+#[config(rename_all = "snake_case")]
+pub struct VllmConfig {
+    /// Environment variable that contains the API key.
+    #[setting(default = "VLLM_API_KEY")]
+    pub api_key_env: String,
+
+    /// The base URL to use for API requests.
+    ///
+    /// The default is `http://127.0.0.1:8000`, which is the default URL for a
+    /// vLLM server.
+    #[setting(default = "http://127.0.0.1:8000")]
+    pub base_url: String,
+}
+
+impl AssignKeyValue for PartialVllmConfig {
+    fn assign(&mut self, kv: KvAssignment) -> AssignResult {
+        match kv.key_string().as_str() {
+            "" => kv.try_merge_object(self)?,
+            "api_key_env" => self.api_key_env = kv.try_some_string()?,
+            "base_url" => self.base_url = kv.try_some_string()?,
+            _ => return missing_key(&kv),
+        }
+
+        Ok(())
+    }
+}
+
+impl PartialConfigDelta for PartialVllmConfig {
+    fn delta(&self, next: Self) -> Self {
+        Self {
+            api_key_env: delta_opt(self.api_key_env.as_ref(), next.api_key_env),
+            base_url: delta_opt(self.base_url.as_ref(), next.base_url),
+        }
+    }
+}
+
+impl FillDefaults for PartialVllmConfig {
+    fn fill_from(self, defaults: Self) -> Self {
+        Self {
+            api_key_env: self.api_key_env.or(defaults.api_key_env),
+            base_url: self.base_url.or(defaults.base_url),
+        }
+    }
+}
+
+impl ToPartial for VllmConfig {
+    fn to_partial(&self) -> Self::Partial {
+        let defaults = Self::Partial::default();
+
+        Self::Partial {
+            api_key_env: partial_opt(&self.api_key_env, defaults.api_key_env),
+            base_url: partial_opt(&self.base_url, defaults.base_url),
+        }
+    }
+}

crates/jp_config/src/snapshots/jp_config__tests__app_config_fields.snap

@@ -54,6 +54,8 @@ expression: "AppConfig::fields()"
     "style.code.line_numbers",
     "providers.mcp",
     "providers.llm.aliases",
+    "providers.llm.vllm.api_key_env",
+    "providers.llm.vllm.base_url",
     "providers.llm.openrouter.api_key_env",
     "providers.llm.openrouter.app_name",
     "providers.llm.openrouter.app_referrer",

crates/jp_config/src/snapshots/jp_config__tests__app_config_schema_shape.snap

@@ -38,7 +38,7 @@ assistant: AssistantConfig
     id: ModelIdOrAliasConfig
       |: ModelIdConfig
         name: string
-        provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+        provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
       |: string
     parameters: ParametersConfig
       max_tokens: int | null
@@ -136,7 +136,7 @@ conversation: ConversationConfig
                   id: ModelIdOrAliasConfig
                     |: ModelIdConfig
                       name: string
-                      provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+                      provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
                     |: string
                   parameters: ParametersConfig
                     max_tokens: int | null
@@ -189,7 +189,7 @@ conversation: ConversationConfig
                     id: ModelIdOrAliasConfig
                       |: ModelIdConfig
                         name: string
-                        provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+                        provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
                       |: string
                     parameters: ParametersConfig
                       max_tokens: int | null
@@ -256,7 +256,7 @@ conversation: ConversationConfig
         id: ModelIdOrAliasConfig
           |: ModelIdConfig
             name: string
-            provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+            provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
           |: string
         parameters: ParametersConfig
           max_tokens: int | null
@@ -363,7 +363,7 @@ conversation: ConversationConfig
           id: ModelIdOrAliasConfig
             |: ModelIdConfig
               name: string
-              provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+              provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
             |: string
           parameters: ParametersConfig
             max_tokens: int | null
@@ -584,7 +584,7 @@ conversation: ConversationConfig
                       |: PartialModelIdOrAliasConfig
                         |: PartialModelIdConfig
                           name?: string | null
-                          provider?: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai" | null
+                          provider?: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai" | null
                         |: string
                       |: null
                     parameters?:
@@ -723,7 +723,7 @@ providers: ProviderConfig
       *: ModelIdOrAliasConfig
         |: ModelIdConfig
           name: string
-          provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+          provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
         |: string
     anthropic: AnthropicConfig
       api_key_env?: string
@@ -753,6 +753,9 @@ providers: ProviderConfig
       app_name?: string
       app_referrer: string | null
       base_url?: string
+    vllm: VllmConfig
+      api_key_env?: string
+      base_url?: string
   mcp:
     *: McpProviderConfig
       |: StdioConfig
@@ -800,7 +803,7 @@ style: StyleConfig
         id: ModelIdOrAliasConfig
           |: ModelIdConfig
             name: string
-            provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "xai"
+            provider: "anthropic" | "cerebras" | "deepseek" | "google" | "llamacpp" | "ollama" | "openai" | "openrouter" | "vllm" | "xai"
           |: string
         parameters: ParametersConfig
           max_tokens: int | null

crates/jp_config/src/snapshots/jp_config__tests__partial_app_config_default.snap

@@ -251,6 +251,10 @@ PartialAppConfig {
                 app_referrer: None,
                 base_url: None,
             },
+            vllm: PartialVllmConfig {
+                api_key_env: None,
+                base_url: None,
+            },
         },
         mcp: {},
     },

crates/jp_config/src/snapshots/jp_config__tests__partial_app_config_default_values.snap

@@ -517,6 +517,14 @@ Ok(
                             "https://openrouter.ai",
                         ),
                     },
+                    vllm: PartialVllmConfig {
+                        api_key_env: Some(
+                            "VLLM_API_KEY",
+                        ),
+                        base_url: Some(
+                            "http://127.0.0.1:8000",
+                        ),
+                    },
                 },
                 mcp: {},
             },

crates/jp_config/src/snapshots/jp_config__tests__partial_app_config_empty_serialize.snap

@@ -251,6 +251,10 @@ PartialAppConfig {
                 app_referrer: None,
                 base_url: None,
             },
+            vllm: PartialVllmConfig {
+                api_key_env: None,
+                base_url: None,
+            },
         },
         mcp: {},
     },

crates/jp_llm/src/provider.rs

@@ -9,6 +9,7 @@ pub mod ollama;
 pub mod openai;
 pub(crate) mod openai_compat;
 pub mod openrouter;
+pub mod vllm;
 
 use std::sync::atomic::{AtomicU64, Ordering};
 
@@ -24,6 +25,7 @@ use llamacpp::Llamacpp;
 use ollama::Ollama;
 use openai::Openai;
 use openrouter::Openrouter;
+use vllm::Vllm;
 
 use crate::{
     error::Result, model::ModelDetails, provider::mock::MockProvider, query::ChatQuery,
@@ -56,6 +58,7 @@ pub fn get_provider(id: ProviderId, config: &LlmProviderConfig) -> Result<Box<dy
         ProviderId::Ollama => Box::new(Ollama::try_from(&config.ollama)?),
         ProviderId::Openai => Box::new(Openai::try_from(&config.openai)?),
         ProviderId::Openrouter => Box::new(Openrouter::try_from(&config.openrouter)?),
+        ProviderId::Vllm => Box::new(Vllm::try_from(&config.vllm)?),
 
         ProviderId::Deepseek => todo!(),
         ProviderId::Xai => todo!(),
@@ -113,6 +116,7 @@ pub(crate) fn build_request_value(
         ProviderId::Openrouter => {
             Openrouter::try_from(&config.openrouter)?.request_value(model, query)
         }
+        ProviderId::Vllm => Vllm::try_from(&config.vllm)?.request_value(model, query),
         ProviderId::Test | ProviderId::Deepseek | ProviderId::Xai => {
             unreachable!("{id:?} is not part of the request snapshot suite")
         }