Add new (basic) agents

docs/openapi/general.yaml

@@ -763,16 +763,8 @@ components:
       required:
       - id
       - full_identity_name
-      - perform_locally
       - model
-      - toolkit_permissions
-      - storage_bucket_permissions
-      - allowed_message_senders
       properties:
-        allowed_message_senders:
-          type: array
-          items:
-            type: string
         api_key:
           type: string
           nullable: true
@@ -785,16 +777,6 @@ components:
           type: string
         model:
           $ref: '#/components/schemas/LLMProviderInterface'
-        perform_locally:
-          type: boolean
-        storage_bucket_permissions:
-          type: array
-          items:
-            type: string
-        toolkit_permissions:
-          type: array
-          items:
-            type: string
     ShinkaiBackend:
       type: object
       required:

docs/openapi/jobs.yaml

@@ -1412,16 +1412,8 @@ components:
       required:
       - id
       - full_identity_name
-      - perform_locally
       - model
-      - toolkit_permissions
-      - storage_bucket_permissions
-      - allowed_message_senders
       properties:
-        allowed_message_senders:
-          type: array
-          items:
-            type: string
         api_key:
           type: string
           nullable: true
@@ -1434,16 +1426,6 @@ components:
           type: string
         model:
           $ref: '#/components/schemas/LLMProviderInterface'
-        perform_locally:
-          type: boolean
-        storage_bucket_permissions:
-          type: array
-          items:
-            type: string
-        toolkit_permissions:
-          type: array
-          items:
-            type: string
     SheetJobAction:
       type: object
       required:

shinkai-bin/shinkai-node/src/llm_provider/error.rs

@@ -87,6 +87,7 @@ pub enum LLMProviderError {
     ToolNotFound(String),
     ToolRetrievalError(String),
     ToolSearchError(String),
+    AgentNotFound(String),
 }
 
 impl fmt::Display for LLMProviderError {
@@ -176,6 +177,7 @@ impl fmt::Display for LLMProviderError {
             LLMProviderError::ToolNotFound(s) => write!(f, "Tool not found: {}", s),
             LLMProviderError::ToolRetrievalError(s) => write!(f, "Tool retrieval error: {}", s),
             LLMProviderError::ToolSearchError(s) => write!(f, "Tool search error: {}", s),
+            LLMProviderError::AgentNotFound(s) => write!(f, "Agent not found: {}", s),
         }
     }
 }
@@ -255,6 +257,7 @@ impl LLMProviderError {
             LLMProviderError::ToolNotFound(_) => "ToolNotFound",
             LLMProviderError::ToolRetrievalError(_) => "ToolRetrievalError",
             LLMProviderError::ToolSearchError(_) => "ToolSearchError",
+            LLMProviderError::AgentNotFound(_) => "AgentNotFound",
         };
 
         let error_message = format!("{}", self);

shinkai-bin/shinkai-node/src/llm_provider/execution/chains/dsl_chain/dsl_inference_chain.rs

@@ -15,7 +15,7 @@ use regex::Regex;
 use shinkai_baml::baml_builder::{BamlConfig, ClientConfig, GeneratorConfig};
 use shinkai_dsl::dsl_schemas::Workflow;
 use shinkai_message_primitives::{
-    schemas::{inbox_name::InboxName, job::JobLike},
+    schemas::{inbox_name::InboxName, job::JobLike, llm_providers::common_agent_llm_provider::ProviderOrAgent},
     shinkai_utils::shinkai_logging::{shinkai_log, ShinkaiLogLevel, ShinkaiLogOption},
 };
 use shinkai_sqlite::logger::{WorkflowLogEntry, WorkflowLogEntryStatus};
@@ -357,6 +357,7 @@ impl AsyncFunction for InferenceFunction {
             },
             None, // this is the config
             self.context.llm_stopper().clone(),
+            self.context.db().clone(),
         )
         .await
         .map_err(|e| WorkflowError::ExecutionError(e.to_string()))?;
@@ -460,6 +461,7 @@ impl AsyncFunction for OpinionatedInferenceFunction {
             },
             None, // this is the config
             self.context.llm_stopper().clone(),
+            self.context.db().clone(),
         )
         .await
         .map_err(|e| WorkflowError::ExecutionError(e.to_string()))?;
@@ -507,11 +509,26 @@ impl AsyncFunction for BamlInference {
 
         // TODO: do we need the job for something?
         // let full_job = self.context.full_job();
-        let llm_provider = self.context.agent();
+        let llm_or_agent_provider = self.context.agent();
 
         let generator_config = GeneratorConfig::default();
 
-        // TODO: add support for other providers
+        let llm_provider = match llm_or_agent_provider {
+            ProviderOrAgent::LLMProvider(provider) => provider,
+            ProviderOrAgent::Agent(agent) => {
+                let llm_provider_id = agent.llm_provider_id.clone();
+                let profile = agent.full_identity_name.clone();
+                let provider = self.context
+                    .db()
+                    .get_llm_provider(&llm_provider_id, &profile)
+                    .map_err(|e| WorkflowError::ExecutionError(e.to_string()))?;
+                &provider
+                    .as_ref()
+                    .ok_or_else(|| WorkflowError::ExecutionError("LLM provider not found".to_string()))?
+                    .clone()
+            }
+        };
+
         let base_url = llm_provider.external_url.clone().unwrap_or_default();
         let base_url = if base_url == "http://localhost:11434" || base_url == "http://localhost:11435" {
             format!("{}/v1", base_url)
@@ -619,7 +636,15 @@ impl AsyncFunction for MultiInferenceFunction {
 
         let mut responses = Vec::new();
         let agent = self.context.agent();
-        let max_tokens = ModelCapabilitiesManager::get_max_input_tokens(&agent.model);
+        let max_tokens = ModelCapabilitiesManager::get_max_input_tokens_for_provider_or_agent(
+            agent.clone(),
+            self.context.db().clone(),
+        );
+
+        let max_tokens = match max_tokens {
+            Some(tokens) => tokens,
+            None => return Err(WorkflowError::ExecutionError("Max tokens not found".to_string())),
+        };
 
         for text in split_texts.iter() {
             let inference_args = vec![

shinkai-bin/shinkai-node/src/llm_provider/execution/chains/dsl_chain/generic_functions.rs

@@ -399,6 +399,7 @@ pub fn search_embeddings_in_job_scope(
 #[cfg(test)]
 mod tests {
     use shinkai_db::db::ShinkaiDB;
+    use shinkai_message_primitives::schemas::llm_providers::common_agent_llm_provider::ProviderOrAgent;
     use shinkai_message_primitives::schemas::llm_providers::serialized_llm_provider::{
         LLMProviderInterface, OpenAI, SerializedLLMProvider,
     };
@@ -696,13 +697,9 @@ mod tests {
         let agent = SerializedLLMProvider {
             id: "test_agent_id".to_string(),
             full_identity_name: agent_name,
-            perform_locally: false,
             external_url: Some("https://api.openai.com".to_string()),
             api_key: Some("mockapikey".to_string()),
             model: LLMProviderInterface::OpenAI(open_ai),
-            toolkit_permissions: vec![],
-            storage_bucket_permissions: vec![],
-            allowed_message_senders: vec![],
         };
         let image_files = HashMap::new();
 
@@ -717,7 +714,7 @@ mod tests {
             },
             None,
             image_files,
-            agent,
+            ProviderOrAgent::LLMProvider(agent),
             HashMap::new(),
             generator,
             ShinkaiName::default_testnet_localhost(),
@@ -832,13 +829,9 @@ mod tests {
         let agent = SerializedLLMProvider {
             id: "test_agent_id_with_query".to_string(),
             full_identity_name: agent_name,
-            perform_locally: false,
             external_url: Some("https://api.openai.com".to_string()),
             api_key: Some("mockapikey".to_string()),
             model: LLMProviderInterface::OpenAI(open_ai),
-            toolkit_permissions: vec![],
-            storage_bucket_permissions: vec![],
-            allowed_message_senders: vec![],
         };
         let image_files = HashMap::new();
 
@@ -853,7 +846,7 @@ mod tests {
             },
             None,
             image_files,
-            agent,
+            ProviderOrAgent::LLMProvider(agent),
             HashMap::new(),
             generator,
             ShinkaiName::default_testnet_localhost(),
@@ -983,13 +976,9 @@ mod tests {
         let agent = SerializedLLMProvider {
             id: "test_agent_id".to_string(),
             full_identity_name: agent_name,
-            perform_locally: false,
             external_url: Some("https://api.openai.com".to_string()),
             api_key: Some("mockapikey".to_string()),
             model: LLMProviderInterface::OpenAI(open_ai),
-            toolkit_permissions: vec![],
-            storage_bucket_permissions: vec![],
-            allowed_message_senders: vec![],
         };
         let image_files = HashMap::new();
 
@@ -1004,7 +993,7 @@ mod tests {
             },
             None,
             image_files,
-            agent,
+            ProviderOrAgent::LLMProvider(agent),
             HashMap::new(),
             generator,
             ShinkaiName::default_testnet_localhost(),

shinkai-bin/shinkai-node/src/llm_provider/execution/chains/dsl_chain/split_text_for_llm.rs

@@ -15,7 +15,12 @@ pub fn split_text_for_llm(
         .clone();
 
     let agent = context.agent();
-    let max_tokens = ModelCapabilitiesManager::get_max_input_tokens(&agent.model);
+    let max_tokens = ModelCapabilitiesManager::get_max_input_tokens_for_provider_or_agent(agent.clone(), context.db().clone());
+
+    let max_tokens = match max_tokens {
+        Some(tokens) => tokens,
+        None => return Err(WorkflowError::ExecutionError("Max tokens not found".to_string())),
+    };
 
     let mut result = Vec::new();
     let current_text = input1.clone();