solana-agent 24.1.1__tar.gz → 25.0.0__tar.gz

{solana_agent-24.1.1 → solana_agent-25.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: solana-agent
-Version: 24.1.1
+Version: 25.0.0
 Summary: Agentic IQ
 License: MIT
 Keywords: ai,openai,ai agents,agi
@@ -14,6 +14,7 @@ Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Dist: instructor (>=1.7.9,<2.0.0)
 Requires-Dist: openai (>=1.71.0,<2.0.0)
 Requires-Dist: pydantic (>=2.11.2,<3.0.0)
 Requires-Dist: pymongo (>=4.11.3,<5.0.0)
@@ -75,16 +76,17 @@ Build your AI business in three lines of code!
 ### Tech
 
 * [Python](https://python.org) - Programming Language
-* [OpenAI](https://openai.com) - LLM Provider
+* [OpenAI](https://openai.com) & [Gemini](https://aistudio.google.com/) - LLM Providers
 * [MongoDB](https://mongodb.com) - Conversational History (optional)
 * [Zep Cloud](https://getzep.com) - Conversational Memory (optional)
 
 ### LLMs
 
-* [gpt-4o-mini](https://platform.openai.com/docs/models/gpt-4o-mini)
+* [gpt-4o-mini](https://platform.openai.com/docs/models/gpt-4o-mini) or [gemini-2.0-flash](https://ai.google.dev/gemini-api/docs/models#gemini-2.0-flash)
 * [gpt-4o-mini-tts](https://platform.openai.com/docs/models/gpt-4o-mini-tts)
 * [gpt-4o-mini-transcribe](https://platform.openai.com/docs/models/gpt-4o-mini-transcribe)
 
+
 ## Installation
 
 You can install Solana Agent using pip:
@@ -144,6 +146,9 @@ Keep this in mind while designing your agentic systems using Solana Agent.
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -167,24 +172,15 @@ async for response in solana_agent.process("user123", "What are the latest AI de
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 1 (agent)
-
-* OpenAI API calls with tool call = 2 (agent, agent)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 2 (router, agent)
-
-* OpenAI API calls with tool call = 3 (router, agent, agent)
-
 ### Audio/Audio Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -210,25 +206,15 @@ async for response in solana_agent.process("user123", audio_content, output_form
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 3 (audio transcribe, agent, TTS)
-
-* OpenAI API calls with tool call = 4 (audio transcribe, agent, agent, TTS)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 4 (router, audio transcribe, agent, TTS)
-
-* OpenAI API calls with tool call = 5 (router, audio transcribe, agent, agent, TTS)
-
-
 ### Text/Audio Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -252,24 +238,15 @@ async for response in solana_agent.process("user123", "What is the latest news o
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 2 (agent, TTS)
-
-* OpenAI API calls with tool call = 3 (agent, agent, TTS)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 3 (router, agent, TTS)
-
-* OpenAI API calls with tool call = 4 (router, agent, agent, TTS)
-
 ### Audio/Text Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -295,18 +272,6 @@ async for response in solana_agent.process("user123", audio_content, audio_input
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 2 (audio transcribe, agent)
-
-* OpenAI API calls with tool call = 3 (audio transcribe, agent, agent)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 3 (router, audio transcribe, agent)
-
-* OpenAI API calls with tool call = 4 (router, audio transcribe, agent, agent)
-
 ## Optional Feature Configs
 
 ### Business Alignment
@@ -332,8 +297,8 @@ config = {
 ```python
 config = {
     "mongo": {
-        "connection_string": "mongodb://localhost:27017",
-        "database": "solana_agent"
+        "connection_string": "your-mongo-connection-string",
+        "database": "your-database-name"
     },
 }
 ```
@@ -348,12 +313,6 @@ config = {
 }
 ```
 
-API Calls:
-
-* Zep adds 2 API calls per user query (GET and POST)
-
-* If the Zep user and session isn't created it creates them for 2 API calls (POST)
-
 ### Customize Speech
 
 This is an audio to audio example using the `audio_instructions` parameter.
@@ -390,6 +349,9 @@ Tools can be used from plugins like Solana Agent Kit (sakit) or via inline tools
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -472,6 +434,9 @@ class TestTool(Tool):
             }
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -511,6 +476,9 @@ This knowledge is accessible to all your AI agents.
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -538,6 +506,9 @@ from solana_agent import SolanaAgent
 from solana_agent.interfaces.services.routing import RoutingService as RoutingServiceInterface
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },

{solana_agent-24.1.1 → solana_agent-25.0.0}/README.md

@@ -51,16 +51,17 @@ Build your AI business in three lines of code!
 ### Tech
 
 * [Python](https://python.org) - Programming Language
-* [OpenAI](https://openai.com) - LLM Provider
+* [OpenAI](https://openai.com) & [Gemini](https://aistudio.google.com/) - LLM Providers
 * [MongoDB](https://mongodb.com) - Conversational History (optional)
 * [Zep Cloud](https://getzep.com) - Conversational Memory (optional)
 
 ### LLMs
 
-* [gpt-4o-mini](https://platform.openai.com/docs/models/gpt-4o-mini)
+* [gpt-4o-mini](https://platform.openai.com/docs/models/gpt-4o-mini) or [gemini-2.0-flash](https://ai.google.dev/gemini-api/docs/models#gemini-2.0-flash)
 * [gpt-4o-mini-tts](https://platform.openai.com/docs/models/gpt-4o-mini-tts)
 * [gpt-4o-mini-transcribe](https://platform.openai.com/docs/models/gpt-4o-mini-transcribe)
 
+
 ## Installation
 
 You can install Solana Agent using pip:
@@ -120,6 +121,9 @@ Keep this in mind while designing your agentic systems using Solana Agent.
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -143,24 +147,15 @@ async for response in solana_agent.process("user123", "What are the latest AI de
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 1 (agent)
-
-* OpenAI API calls with tool call = 2 (agent, agent)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 2 (router, agent)
-
-* OpenAI API calls with tool call = 3 (router, agent, agent)
-
 ### Audio/Audio Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -186,25 +181,15 @@ async for response in solana_agent.process("user123", audio_content, output_form
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 3 (audio transcribe, agent, TTS)
-
-* OpenAI API calls with tool call = 4 (audio transcribe, agent, agent, TTS)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 4 (router, audio transcribe, agent, TTS)
-
-* OpenAI API calls with tool call = 5 (router, audio transcribe, agent, agent, TTS)
-
-
 ### Text/Audio Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -228,24 +213,15 @@ async for response in solana_agent.process("user123", "What is the latest news o
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 2 (agent, TTS)
-
-* OpenAI API calls with tool call = 3 (agent, agent, TTS)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 3 (router, agent, TTS)
-
-* OpenAI API calls with tool call = 4 (router, agent, agent, TTS)
-
 ### Audio/Text Streaming
 
 ```python
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -271,18 +247,6 @@ async for response in solana_agent.process("user123", audio_content, audio_input
     print(response, end="")
 ```
 
-Single Agent:
-
-* OpenAI API calls with no tool call = 2 (audio transcribe, agent)
-
-* OpenAI API calls with tool call = 3 (audio transcribe, agent, agent)
-
-Multiple Agents:
-
-* OpenAI API calls with no tool call = 3 (router, audio transcribe, agent)
-
-* OpenAI API calls with tool call = 4 (router, audio transcribe, agent, agent)
-
 ## Optional Feature Configs
 
 ### Business Alignment
@@ -308,8 +272,8 @@ config = {
 ```python
 config = {
     "mongo": {
-        "connection_string": "mongodb://localhost:27017",
-        "database": "solana_agent"
+        "connection_string": "your-mongo-connection-string",
+        "database": "your-database-name"
     },
 }
 ```
@@ -324,12 +288,6 @@ config = {
 }
 ```
 
-API Calls:
-
-* Zep adds 2 API calls per user query (GET and POST)
-
-* If the Zep user and session isn't created it creates them for 2 API calls (POST)
-
 ### Customize Speech
 
 This is an audio to audio example using the `audio_instructions` parameter.
@@ -366,6 +324,9 @@ Tools can be used from plugins like Solana Agent Kit (sakit) or via inline tools
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -448,6 +409,9 @@ class TestTool(Tool):
             }
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -487,6 +451,9 @@ This knowledge is accessible to all your AI agents.
 from solana_agent import SolanaAgent
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },
@@ -514,6 +481,9 @@ from solana_agent import SolanaAgent
 from solana_agent.interfaces.services.routing import RoutingService as RoutingServiceInterface
 
 config = {
+    "gemini": {
+        "api_key": "your-gemini-api-key",
+    },
     "openai": {
         "api_key": "your-openai-api-key",
     },

{solana_agent-24.1.1 → solana_agent-25.0.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "solana-agent"
-version = "24.1.1"
+version = "25.0.0"
 description = "Agentic IQ"
 authors = ["Bevan Hunt <bevan@bevanhunt.com>"]
 license = "MIT"
@@ -27,6 +27,7 @@ openai = "^1.71.0"
 pydantic = "^2.11.2"
 pymongo = "^4.11.3"
 zep-cloud = "^2.9.0"
+instructor = "^1.7.9"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^8.3.5"

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/adapters/llm_adapter.py

@@ -3,10 +3,12 @@ LLM provider adapters for the Solana Agent system.
 
 These adapters implement the LLMProvider interface for different LLM services.
 """
-from typing import AsyncGenerator, Literal, Type, TypeVar
+from typing import AsyncGenerator, Literal, Optional, Type, TypeVar
 
 from openai import AsyncOpenAI
 from pydantic import BaseModel
+import instructor
+from instructor import Mode
 
 from solana_agent.interfaces.providers.llm import LLMProvider
 
@@ -103,6 +105,9 @@ class OpenAIAdapter(LLMProvider):
         self,
         prompt: str,
         system_prompt: str = "",
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> AsyncGenerator[str, None]:  # pragma: no cover
         """Generate text from OpenAI models."""
         messages = []
@@ -118,8 +123,18 @@ class OpenAIAdapter(LLMProvider):
             "stream": True,
             "model": self.text_model,
         }
+
+        client = self.client
+
+        if api_key and base_url:
+            client.api_key = api_key
+            client.base_url = base_url
+
+        if model:
+            request_params["model"] = model
+
         try:
-            response = await self.client.chat.completions.create(**request_params)
+            response = await client.chat.completions.create(**request_params)
 
             async for chunk in response:
                 if chunk.choices:
@@ -138,21 +153,83 @@ class OpenAIAdapter(LLMProvider):
         prompt: str,
         system_prompt: str,
         model_class: Type[T],
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> T:  # pragma: no cover
-        """Generate structured output using Pydantic model parsing."""
-        messages = []
-        if system_prompt:
-            messages.append({"role": "system", "content": system_prompt})
+        """Generate structured output using Pydantic model parsing with Instructor."""
 
+        messages = []
+        messages.append({"role": "system", "content": system_prompt})
         messages.append({"role": "user", "content": prompt})
 
         try:
-            # First try the beta parsing API
-            completion = await self.client.beta.chat.completions.parse(
+            if api_key and base_url:
+                self.client.api_key = api_key
+                self.client.base_url = base_url
+
+            if model:
+                self.parse_model = model
+
+            # Create a patched client with TOOLS_STRICT mode
+            patched_client = instructor.from_openai(
+                self.client, mode=Mode.TOOLS_STRICT)
+
+            # Use instructor's structured generation with function calling
+            response = await patched_client.chat.completions.create(
                 model=self.parse_model,
                 messages=messages,
-                response_format=model_class,
+                response_model=model_class,
+                max_retries=2  # Automatically retry on validation errors
             )
-            return completion.choices[0].message.parsed
+            return response
         except Exception as e:
-            print(f"Error with beta.parse method: {e}")
+            print(
+                f"Error with instructor parsing (TOOLS_STRICT mode): {e}")
+
+            try:
+                # First fallback: Try regular JSON mode
+                patched_client = instructor.from_openai(
+                    self.client, mode=Mode.JSON)
+                response = await patched_client.chat.completions.create(
+                    model=self.parse_model,
+                    messages=messages,
+                    response_model=model_class,
+                    max_retries=1
+                )
+                return response
+            except Exception as json_error:
+                print(f"JSON mode fallback also failed: {json_error}")
+
+                try:
+                    # Final fallback: Manual extraction with a detailed prompt
+                    fallback_system_prompt = f"""
+                    {system_prompt}
+
+                    You must respond with valid JSON that can be parsed as the following Pydantic model:
+                    {model_class.model_json_schema()}
+
+                    Ensure the response contains ONLY the JSON object and nothing else.
+                    """
+
+                    # Regular completion without instructor
+                    completion = await self.client.chat.completions.create(
+                        model=self.parse_model,
+                        messages=[
+                            {"role": "system", "content": fallback_system_prompt},
+                            {"role": "user", "content": prompt}
+                        ],
+                        response_format={"type": "json_object"}
+                    )
+
+                    # Extract and parse the JSON response
+                    json_str = completion.choices[0].message.content
+
+                    # Use Pydantic to parse and validate
+                    return model_class.model_validate_json(json_str)
+
+                except Exception as fallback_error:
+                    print(f"All fallback methods failed: {fallback_error}")
+                    raise ValueError(
+                        f"Failed to generate structured output: {e}. All fallbacks failed."
+                    ) from e

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/factories/agent_factory.py

@@ -86,23 +86,43 @@ class SolanaAgentFactory:
                 zep_api_key=config["zep"].get("api_key")
             )
 
-        # Create primary services
-        agent_service = AgentService(
-            llm_provider=llm_adapter,
-            business_mission=business_mission,
-            config=config,
-        )
+        if "gemini" in config and "api_key" in config["gemini"]:
+            # Create primary services
+            agent_service = AgentService(
+                llm_provider=llm_adapter,
+                business_mission=business_mission,
+                config=config,
+                api_key=config["gemini"]["api_key"],
+                base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
+                model="gemini-2.0-flash",
+            )
+
+            # Create routing service
+            routing_service = RoutingService(
+                llm_provider=llm_adapter,
+                agent_service=agent_service,
+                api_key=config["gemini"]["api_key"],
+                base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
+                model="gemini-2.0-flash",
+            )
+        else:
+            # Create primary services
+            agent_service = AgentService(
+                llm_provider=llm_adapter,
+                business_mission=business_mission,
+                config=config,
+            )
+
+            # Create routing service
+            routing_service = RoutingService(
+                llm_provider=llm_adapter,
+                agent_service=agent_service,
+            )
 
         # Debug the agent service tool registry
         print(
             f"Agent service tools after initialization: {agent_service.tool_registry.list_all_tools()}")
 
-        # Create routing service
-        routing_service = RoutingService(
-            llm_provider=llm_adapter,
-            agent_service=agent_service,
-        )
-
         # Initialize plugin system
         agent_service.plugin_manager = PluginManager(
             config=config,

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/interfaces/providers/llm.py

@@ -15,13 +15,21 @@ class LLMProvider(ABC):
         self,
         prompt: str,
         system_prompt: str = "",
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None,
     ) -> AsyncGenerator[str, None]:
         """Generate text from the language model."""
         pass
 
     @abstractmethod
     async def parse_structured_output(
-        self, prompt: str, system_prompt: str, model_class: Type[T],
+        self, prompt: str,
+        system_prompt: str,
+        model_class: Type[T],
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None
     ) -> T:
         """Generate structured output using a specific model class."""
         pass
@@ -49,13 +57,3 @@ class LLMProvider(ABC):
     ) -> AsyncGenerator[str, None]:
         """Transcribe audio from the language model."""
         pass
-
-    @abstractmethod
-    async def realtime_audio_transcription(
-        self,
-        audio_generator: AsyncGenerator[bytes, None],
-        transcription_config: Optional[Dict[str, Any]] = None,
-        on_event: Optional[Callable[[Dict[str, Any]], Any]] = None,
-    ) -> AsyncGenerator[str, None]:
-        """Stream real-time audio transcription from the language model."""
-        pass

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/interfaces/services/agent.py

@@ -30,9 +30,6 @@ class AgentService(ABC):
         audio_instructions: str = "You speak in a friendly and helpful manner.",
         audio_output_format: Literal['mp3', 'opus',
                                      'aac', 'flac', 'wav', 'pcm'] = "aac",
-        audio_input_format: Literal[
-            "flac", "mp3", "mp4", "mpeg", "mpga", "m4a", "ogg", "wav", "webm"
-        ] = "mp4",
         prompt: Optional[str] = None,
     ) -> AsyncGenerator[Union[str, bytes], None]:
         """Generate a response from an agent."""

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/services/agent.py

@@ -26,6 +26,9 @@ class AgentService(AgentServiceInterface):
         llm_provider: LLMProvider,
         business_mission: Optional[BusinessMission] = None,
         config: Optional[Dict[str, Any]] = None,
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None,
     ):
         """Initialize the agent service.
 
@@ -40,6 +43,9 @@ class AgentService(AgentServiceInterface):
         self.last_text_response = ""
         self.tool_registry = ToolRegistry(config=self.config)
         self.agents: List[AIAgent] = []
+        self.api_key = api_key
+        self.base_url = base_url
+        self.model = model
 
         self.plugin_manager = PluginManager(
             config=self.config,
@@ -173,9 +179,6 @@ class AgentService(AgentServiceInterface):
         audio_instructions: str = "You speak in a friendly and helpful manner.",
         audio_output_format: Literal['mp3', 'opus',
                                      'aac', 'flac', 'wav', 'pcm'] = "aac",
-        audio_input_format: Literal[
-            "flac", "mp3", "mp4", "mpeg", "mpga", "m4a", "ogg", "wav", "webm"
-        ] = "mp4",
         prompt: Optional[str] = None,
     ) -> AsyncGenerator[Union[str, bytes], None]:  # pragma: no cover
         """Generate a response with support for text/audio input/output."""
@@ -191,14 +194,6 @@ class AgentService(AgentServiceInterface):
             return
 
         try:
-            # Handle audio input if provided - KEEP REAL-TIME AUDIO TRANSCRIPTION
-            query_text = ""
-            if not isinstance(query, str):
-                async for transcript in self.llm_provider.transcribe_audio(query, input_format=audio_input_format):
-                    query_text += transcript
-            else:
-                query_text = query
-
             # Get system prompt
             system_prompt = self.get_agent_system_prompt(agent_name)
 
@@ -234,10 +229,13 @@ class AgentService(AgentServiceInterface):
 
             # Generate and stream response (ALWAYS use non-realtime for text generation)
             print(
-                f"Generating response with {len(query_text)} characters of query text")
+                f"Generating response with {len(query)} characters of query text")
             async for chunk in self.llm_provider.generate_text(
-                prompt=query_text,
+                prompt=query,
                 system_prompt=tool_calling_system_prompt,
+                api_key=self.api_key,
+                base_url=self.base_url,
+                model=self.model,
             ):
                 # If we have pending text from the previous chunk, combine it with this chunk
                 if pending_chunk:
@@ -288,7 +286,7 @@ class AgentService(AgentServiceInterface):
 
                         # Create new prompt with search/tool results
                         # Using "Search Result" instead of "TOOL RESPONSE" to avoid model repeating "TOOL"
-                        user_prompt = f"{query_text}\n\nSearch Result: {response_text}"
+                        user_prompt = f"{query}\n\nSearch Result: {response_text}"
                         tool_system_prompt = system_prompt + \
                             "\n DO NOT use the tool calling format again."
 
@@ -299,6 +297,9 @@ class AgentService(AgentServiceInterface):
                             async for processed_chunk in self.llm_provider.generate_text(
                                 prompt=user_prompt,
                                 system_prompt=tool_system_prompt,
+                                api_key=self.api_key,
+                                base_url=self.base_url,
+                                model=self.model,
                             ):
                                 complete_text_response += processed_chunk
                                 yield processed_chunk

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/services/query.py

@@ -112,11 +112,10 @@ class QueryService(QueryServiceInterface):
                 async for audio_chunk in self.agent_service.generate_response(
                     agent_name=agent_name,
                     user_id=user_id,
-                    query=query,
+                    query=user_text,
                     memory_context=memory_context,
                     output_format="audio",
                     audio_voice=audio_voice,
-                    audio_input_format=audio_input_format,
                     audio_output_format=audio_output_format,
                     audio_instructions=audio_instructions,
                     prompt=prompt,

{solana_agent-24.1.1 → solana_agent-25.0.0}/solana_agent/services/routing.py

@@ -18,6 +18,9 @@ class RoutingService(RoutingServiceInterface):
         self,
         llm_provider: LLMProvider,
         agent_service: AgentService,
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+        model: Optional[str] = None,
     ):
         """Initialize the routing service.
 
@@ -27,6 +30,9 @@ class RoutingService(RoutingServiceInterface):
         """
         self.llm_provider = llm_provider
         self.agent_service = agent_service
+        self.api_key = api_key
+        self.base_url = base_url
+        self.model = model
 
     async def _analyze_query(self, query: str) -> Dict[str, Any]:
         """Analyze a query to determine routing information.
@@ -75,6 +81,9 @@ class RoutingService(RoutingServiceInterface):
                 prompt=prompt,
                 system_prompt="Match user queries to the most appropriate agent based on specializations.",
                 model_class=QueryAnalysis,
+                api_key=self.api_key,
+                base_url=self.base_url,
+                model=self.model,
             )
 
             return {