autonomous-app 0.3.27__py3-none-any.whl → 0.3.29__py3-none-any.whl

autonomous/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.3.27"
+__version__ = "0.3.29"
 
 from dotenv import load_dotenv
 

autonomous/ai/baseagent.py

@@ -3,24 +3,16 @@ from autonomous.model.autoattr import ReferenceAttr
 from autonomous.model.automodel import AutoModel
 
 from .models.gemini import GeminiAIModel
+from .models.local_model import LocalAIModel
 from .models.openai import OpenAIModel
 
 
-def clear_agents():
-    for agent in OpenAIModel.all():
-        log(f"Deleting {agent.name}")
-        agent.clear_agents()
-        agent.clear_files()
-        agent.delete()
-    return "Success"
-
-
 class BaseAgent(AutoModel):
     meta = {"abstract": True, "allow_inheritance": True, "strict": False}
 
-    client = ReferenceAttr(choices=[GeminiAIModel])
+    client = ReferenceAttr(choices=[LocalAIModel])
 
-    _ai_model = GeminiAIModel
+    _ai_model = LocalAIModel
 
     def delete(self):
         if self.client:

autonomous/ai/jsonagent.py

@@ -1,8 +1,9 @@
 import json
 
+from autonomous.ai.baseagent import BaseAgent
 from autonomous.model.autoattr import ReferenceAttr, StringAttr
 from autonomous.model.automodel import AutoModel
-from autonomous.ai.baseagent import BaseAgent
+
 from .models.openai import OpenAIModel
 
 
@@ -24,3 +25,6 @@ class JSONAgent(BaseAgent):
         elif not isinstance(result, dict):
             raise ValueError(f"Invalid JSON response from AI model.\n\n{result}")
         return result
+
+    def upload(self, file):
+        return self.get_client().upload(file=file)

autonomous/ai/models/gemini.py

@@ -23,6 +23,16 @@ class GeminiAIModel(AutoModel):
     _json_model = "gemini-3-pro-preview"
     _stt_model = "gemini-3-pro-preview"
     _tts_model = "gemini-2.5-flash-preview-tts"
+
+    messages = ListAttr(StringAttr(default=[]))
+    name = StringAttr(default="agent")
+    instructions = StringAttr(
+        default="You are highly skilled AI trained to assist with various tasks."
+    )
+    description = StringAttr(
+        default="A helpful AI assistant trained to assist with various tasks."
+    )
+
     MAX_FILES = 14
     MAX_SUMMARY_TOKEN_LENGTH = 10000
     VOICES = {
@@ -58,15 +68,6 @@ class GeminiAIModel(AutoModel):
         "Sulafar": ["female"],
     }
 
-    messages = ListAttr(StringAttr(default=[]))
-    name = StringAttr(default="agent")
-    instructions = StringAttr(
-        default="You are highly skilled AI trained to assist with various tasks."
-    )
-    description = StringAttr(
-        default="A helpful AI assistant trained to assist with various tasks."
-    )
-
     @property
     def client(self):
         if not self._client:
@@ -114,8 +115,16 @@ class GeminiAIModel(AutoModel):
         existing_files = self.client.files.list()
         log(f"Existing files: {[f.display_name for f in existing_files]}", _print=True)
         for f in existing_files:
-            result = self.client.files.delete(name=f.name)
-            log(f"Deleting old version of {f.name}: {result}", _print=True)
+            # Delete old files (older than 10 minutes)
+            age_seconds = (
+                (time.time() - f.update_time.timestamp())
+                if f.update_time
+                else (time.time() - f.create_time.timestamp())
+            )
+            log(age_seconds, _print=True)
+            if age_seconds > 900:
+                result = self.client.files.delete(name=f.name)
+                log(f"Deleting old version of {f.name}: {result}", _print=True)
         file_refs = []
         for file_dict in file_list:
             fn = file_dict["name"]
@@ -132,7 +141,10 @@ class GeminiAIModel(AutoModel):
                 time.sleep(1)
                 uploaded_file = self.client.get_file(uploaded_file.name)
             file_refs.append(uploaded_file)
-        return file_refs
+            return file_refs
+
+    def upload(self, file):
+        return self._add_files([file])
 
     def generate_json(self, message, function, additional_instructions="", **kwargs):
         # The API call must use the 'tools' parameter instead of 'response_json_schema'

autonomous/ai/models/local_model.py

@@ -0,0 +1,272 @@
+import io
+import json
+import os
+import random
+import re
+import wave
+
+import requests
+from pydub import AudioSegment
+
+from autonomous import log
+from autonomous.model.autoattr import ListAttr, StringAttr
+from autonomous.model.automodel import AutoModel
+
+
+class LocalAIModel(AutoModel):
+    # Configuration
+    _ollama_url = os.environ.get("OLLAMA_API_BASE", "http://ollama_internal:11434/api")
+    _media_url = os.environ.get("MEDIA_API_BASE", "http://media_ai_internal:5005")
+
+    # Models to use in Ollama
+    _text_model = "mistral-nemo"
+    _json_model = "mistral-nemo"
+
+    messages = ListAttr(StringAttr(default=[]))
+    name = StringAttr(default="agent")
+    instructions = StringAttr(
+        default="You are highly skilled AI trained to assist with various tasks."
+    )
+
+    # Keep your voice list (mapped to random seeds/embeddings in the future)
+    VOICES = {
+        "Zephyr": ["female"],
+        "Puck": ["male"],
+        "Charon": ["male"],
+        "Kore": ["female"],
+        "Fenrir": ["non-binary"],
+        "Leda": ["female"],
+        "Orus": ["male"],
+        "Aoede": ["female"],
+        "Callirhoe": ["female"],
+        "Autonoe": ["female"],
+        "Enceladus": ["male"],
+        "Iapetus": ["male"],
+        "Umbriel": ["male"],
+        "Algieba": ["male"],
+        "Despina": ["female"],
+        "Erinome": ["female"],
+        "Algenib": ["male"],
+        "Rasalgethi": ["non-binary"],
+        "Laomedeia": ["female"],
+        "Achernar": ["female"],
+        "Alnilam": ["male"],
+        "Schedar": ["male"],
+        "Gacrux": ["female"],
+        "Pulcherrima": ["non-binary"],
+        "Achird": ["male"],
+        "Zubenelgenubi": ["male"],
+        "Vindemiatrix": ["female"],
+        "Sadachbia": ["male"],
+        "Sadaltager": ["male"],
+        "Sulafar": ["female"],
+    }
+
+    def _convert_tools_to_json_schema(self, user_function):
+        """
+        Ollama doesn't support 'tools' strictly yet.
+        We convert the tool definition into a system prompt instruction.
+        """
+        schema = {
+            "name": user_function.get("name"),
+            "parameters": user_function.get("parameters"),
+        }
+        return json.dumps(schema, indent=2)
+
+    def generate_json(self, message, function, additional_instructions="", **kwargs):
+        """
+        Mimics Gemini's tool use by forcing Ollama into JSON mode
+        and injecting the schema into the prompt.
+        """
+        schema_str = self._convert_tools_to_json_schema(function)
+
+        system_prompt = (
+            f"{self.instructions}. {additional_instructions}\n"
+            f"You must respond strictly with a valid JSON object matching this schema:\n"
+            f"{schema_str}\n"
+            f"Do not include markdown formatting or explanations."
+        )
+
+        payload = {
+            "model": self._json_model,
+            "prompt": message,
+            "system": system_prompt,
+            "format": "json",  # Force JSON mode
+            "stream": False,
+        }
+
+        try:
+            response = requests.post(f"{self._ollama_url}/generate", json=payload)
+            response.raise_for_status()
+            result_text = response.json().get("response", "{}")
+
+            # log(f"Raw Local JSON: {result_text}", _print=True)
+            return json.loads(result_text)
+
+        except Exception as e:
+            log(f"==== LocalAI JSON Error: {e} ====", _print=True)
+            return {}
+
+    def generate_text(self, message, additional_instructions="", **kwargs):
+        """
+        Standard text generation via Ollama.
+        """
+        payload = {
+            "model": self._text_model,
+            "prompt": message,
+            "system": f"{self.instructions}. {additional_instructions}",
+            "stream": False,
+        }
+
+        # Handle 'files' (Ollama supports images in base64, but not arbitrary files easily yet)
+        # If files are text, you should read them and append to prompt.
+        if file_list := kwargs.get("files"):
+            for file_dict in file_list:
+                fn = file_dict["name"]
+                fileobj = file_dict["file"]
+                if fn.lower().endswith((".txt", ".md", ".json", ".csv")):
+                    content = fileobj.read()
+                    if isinstance(content, bytes):
+                        content = content.decode("utf-8", errors="ignore")
+                    payload["prompt"] += f"\n\nContents of {fn}:\n{content}"
+
+        try:
+            response = requests.post(f"{self._ollama_url}/generate", json=payload)
+            response.raise_for_status()
+            return response.json().get("response", "")
+        except Exception as e:
+            log(f"==== LocalAI Text Error: {e} ====", _print=True)
+            return "Error generating text."
+
+    def summarize_text(self, text, primer="", **kwargs):
+        primer = primer or "Summarize the following text concisely."
+
+        # Simple chunking logic (similar to your original)
+        # Note: Mistral-Nemo has a large context window (128k), so chunking
+        # is less necessary than with older models, but we keep it for safety.
+        max_chars = 12000  # Roughly 3k tokens
+        chunks = [text[i : i + max_chars] for i in range(0, len(text), max_chars)]
+
+        full_summary = ""
+        for chunk in chunks:
+            payload = {
+                "model": self._text_model,
+                "prompt": f"{primer}:\n\n{chunk}",
+                "stream": False,
+            }
+            try:
+                res = requests.post(f"{self._ollama_url}/generate", json=payload)
+                full_summary += res.json().get("response", "") + "\n"
+            except Exception as e:
+                log(f"Summary Error: {e}", _print=True)
+                break
+
+        return full_summary
+
+    def generate_audio_text(self, audio_file, prompt="", **kwargs):
+        """
+        Sends audio bytes to the Media AI container for Whisper transcription.
+        """
+        try:
+            # Prepare the file for upload
+            # audio_file is likely bytes, so we wrap in BytesIO if needed
+            if isinstance(audio_file, bytes):
+                f_obj = io.BytesIO(audio_file)
+            else:
+                f_obj = audio_file
+
+            files = {"file": ("audio.mp3", f_obj, "audio/mpeg")}
+
+            response = requests.post(f"{self._media_url}/transcribe", files=files)
+            response.raise_for_status()
+            return response.json().get("text", "")
+
+        except Exception as e:
+            log(f"STT Error: {e}", _print=True)
+            return ""
+
+    def generate_audio(self, prompt, voice=None, **kwargs):
+        """
+        Sends text to the Media AI container for TTS.
+        """
+        voice = voice or random.choice(list(self.VOICES.keys()))
+
+        try:
+            payload = {"text": prompt, "voice": voice}
+            response = requests.post(f"{self._media_url}/tts", json=payload)
+            response.raise_for_status()
+
+            # Response content is WAV bytes
+            wav_bytes = response.content
+
+            # Convert to MP3 to match your original interface (using pydub)
+            audio = AudioSegment.from_file(io.BytesIO(wav_bytes), format="wav")
+            mp3_buffer = io.BytesIO()
+            audio.export(mp3_buffer, format="mp3")
+            return mp3_buffer.getvalue()
+
+        except Exception as e:
+            log(f"TTS Error: {e}", _print=True)
+            return None
+
+    def generate_image(self, prompt, **kwargs):
+        """
+        Generates an image using Local AI.
+        If 'files' are provided, performs Image-to-Image generation using the first file as reference.
+        """
+        try:
+            # Prepare the multipart data
+            # We send the prompt as a form field
+            data = {"prompt": prompt}
+            files = {}
+
+            # Check if reference images were passed
+            if kwargs.get("files"):
+                # Take the first available file
+                for fn, f_bytes in kwargs.get("files").items():
+                    # If f_bytes is bytes, wrap in IO, else assume it's file-like
+                    if isinstance(f_bytes, bytes):
+                        file_obj = io.BytesIO(f_bytes)
+                    else:
+                        file_obj = f_bytes
+
+                    # Add to the request files
+                    # Key must be 'file' to match server.py logic
+                    files["file"] = (fn, file_obj, "image/png")
+                    break  # We only support 1 reference image for SD Img2Img
+
+            # Send Request
+            if files:
+                # Multipart/form-data request (Prompt + File)
+                response = requests.post(
+                    f"{self._media_url}/generate-image", data=data, files=files
+                )
+            else:
+                # Standard request (Prompt only) - server.py handles request.form vs json
+                # But our updated server expects form data for consistency
+                response = requests.post(f"{self._media_url}/generate-image", data=data)
+
+            response.raise_for_status()
+
+            # Returns WebP bytes directly
+            return response.content
+
+        except Exception as e:
+            log(f"Image Gen Error: {e}", _print=True)
+            return None
+
+    def list_voices(self, filters=[]):
+        # Same logic as before
+        if not filters:
+            return list(self.VOICES.keys())
+        voices = []
+        for voice, attribs in self.VOICES.items():
+            if any(f.lower() in attribs for f in filters):
+                voices.append(voice)
+        return voices
+
+    # Unused methods from original that don't apply to Local AI
+    def upload(self, file):
+        # Local models don't really have a "File Store" API like Gemini.
+        # We handle context by passing text directly in prompt.
+        pass

{autonomous_app-0.3.27.dist-info → autonomous_app-0.3.29.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: autonomous-app
-Version: 0.3.27
+Version: 0.3.29
 Summary: Containerized application framework built on Flask with additional libraries and tools for rapid development of web applications.
 Author-email: Steven A Moore <samoore@binghamton.edu>
 Project-URL: homepage, https://github.com/Sallenmoore/autonomous

{autonomous_app-0.3.27.dist-info → autonomous_app-0.3.29.dist-info}/RECORD

@@ -1,17 +1,17 @@
-autonomous/__init__.py,sha256=kn0Y6qhJcvWmWzx8A45mzDbvKLy0PoPhcEa4xeWifcU,95
+autonomous/__init__.py,sha256=j-9rQTP_Ejh4XtrStDlXn9R3weOnMognFa40DRWU_V0,95
 autonomous/cli.py,sha256=z4AaGeWNW_uBLFAHng0J_lfS9v3fXemK1PeT85u4Eo4,42
 autonomous/logger.py,sha256=NQtgEaTWNAWfLSgqSP7ksXj1GpOuCgoUV711kSMm-WA,2022
 autonomous/ai/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autonomous/ai/audioagent.py,sha256=SvPLzKgqUnrkcsR7y93aURSrStIrryuntQMPS1SzUXw,1033
-autonomous/ai/baseagent.py,sha256=HYCqC4HmK5afNMunmTkhRE8O0OaONl2GxXnISkdOM58,1094
+autonomous/ai/baseagent.py,sha256=sJDIrCzhUp1OwkfyRqT0ZaB1fk5pV8K5L3jaTxzi9DI,940
 autonomous/ai/imageagent.py,sha256=bIOrgg_CM-rgfyLme7V9vPqP8WKVMIAVoB2E9lLtIRk,521
-autonomous/ai/jsonagent.py,sha256=ldfWHtKfLa2ypoM95U6PFETAE9R5B53s5oGzIzF7dQk,984
+autonomous/ai/jsonagent.py,sha256=VQGhK0RFo0H_eVH9dAyf4_lp-RIpdgH988joLoKjm94,1065
 autonomous/ai/textagent.py,sha256=1yM1aMvws64PocvG_L-POMDKjxq2JDuGqgc3haUHybU,926
 autonomous/ai/models/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autonomous/ai/models/aws.py,sha256=bGDjnGTm350zOqor9IsICzUkBUN2bubGI_ZssQuSXIw,12715
 autonomous/ai/models/deepseek.py,sha256=fkoi-hJp60yFlZ9Cb9PdUrmNSErYltQ5ezkUI75llXc,2734
-autonomous/ai/models/gemini.py,sha256=jTCOEoCJd-sCB1oPjYuCitFuIheut3RvN4DpuGabx0c,13839
-autonomous/ai/models/local.py,sha256=fkoi-hJp60yFlZ9Cb9PdUrmNSErYltQ5ezkUI75llXc,2734
+autonomous/ai/models/gemini.py,sha256=jrTMbh8SAdzzz27elOhs82iwjyutYcy8fvTOSdW-GFQ,14247
+autonomous/ai/models/local_model.py,sha256=GBXUelGUObo33BYaPVbCqI0asYaFLET2JNMFSXqxngw,9846
 autonomous/ai/models/openai.py,sha256=2-LttCm6woGklaLbs1H5LjlbfM-7leDwGmC9vksSqW4,13135
 autonomous/apis/version_control/GHCallbacks.py,sha256=AyiUlYfV5JePi11GVyqYyXoj5UTbPKzS-HRRI94rjJo,1069
 autonomous/apis/version_control/GHOrganization.py,sha256=mi2livdsGurKiifbvuLwiFbdDzL77IlEfhwEa-tG77I,1155
@@ -57,7 +57,7 @@ autonomous/storage/localstorage.py,sha256=FzrR6O9mMGAZt5dDgqzkeOQVfGRXCygR0kksz2
 autonomous/tasks/__init__.py,sha256=pn7iZ14MhcHUdzcLkfkd4-45wgPP0tXahAz_cFgb_Tg,32
 autonomous/tasks/autotask.py,sha256=aK5iapDhgcAic3F5ZYMAhNKJkOepj8yWwbMizKDzUwQ,4153
 autonomous/utils/markdown.py,sha256=tf8vlHARiQO1X_aGbqlYozzP_TbdiDRT9EEP6aFRQo0,2153
-autonomous_app-0.3.27.dist-info/METADATA,sha256=x9PPnKHTy5dMQks_32hax4zXjUk1dxAAmlbpcG-JpIQ,3015
-autonomous_app-0.3.27.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-autonomous_app-0.3.27.dist-info/top_level.txt,sha256=ZyxWWDdbvZekF3UFunxl4BQsVDb_FOW3eTn0vun_jb4,11
-autonomous_app-0.3.27.dist-info/RECORD,,
+autonomous_app-0.3.29.dist-info/METADATA,sha256=VEPJShGjhIrEukUwiMTlIDiu7bbkEnrvoy2r03WBsQo,3015
+autonomous_app-0.3.29.dist-info/WHEEL,sha256=qELbo2s1Yzl39ZmrAibXA2jjPLUYfnVhUNTlyF1rq0Y,92
+autonomous_app-0.3.29.dist-info/top_level.txt,sha256=ZyxWWDdbvZekF3UFunxl4BQsVDb_FOW3eTn0vun_jb4,11
+autonomous_app-0.3.29.dist-info/RECORD,,

{autonomous_app-0.3.27.dist-info → autonomous_app-0.3.29.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.9.0)
+Generator: setuptools (80.10.1)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

autonomous/ai/models/local.py

@@ -1,99 +0,0 @@
-import io
-import json
-import os
-import random
-import time
-from base64 import b64decode
-
-import openai
-from ollama import ChatResponse, chat
-
-from autonomous import log
-from autonomous.model.autoattr import DictAttr, ListAttr, StringAttr
-from autonomous.model.automodel import AutoModel
-
-
-class LocalAIModel(AutoModel):
-    _client = None
-    instructions = StringAttr(
-        default="You are highly skilled AI trained to assist with various tasks."
-    )
-    description = StringAttr(
-        default="A helpful AI assistant trained to assist with various tasks."
-    )
-
-    @property
-    def client(self):
-        if not self._client:
-            self._client = "deepseek-r1"  # OpenAI(api_key=os.environ.get("OPENAI_KEY"))
-        return self._client
-
-    def clear_agent(self):
-        pass
-
-    def clear_agents(self):
-        pass
-
-        # def _get_agent_id(self):
-        #     pass
-
-        # def _add_function(self, user_function):
-        pass
-
-    def _format_messages(self, messages):
-        pass
-
-    def clear_files(self, file_id=None):
-        pass
-
-    def attach_file(self, file_contents, filename="dbdata.json"):
-        pass
-
-    def generate_json(self, messages, function, additional_instructions=""):
-        message = messages + additional_instructions
-        message += f"""
-IMPORTANT: Respond in JSON FORMAT using the SCHEMA below. DO NOT add any text to the response outside of the supplied JSON schema:
-{function}
-"""
-        response: ChatResponse = chat(
-            model=self.client,
-            messages=[
-                {
-                    "role": "user",
-                    "content": message,
-                },
-            ],
-        )
-        return response.message.content
-
-    def generate_text(self, messages, additional_instructions=""):
-        message = messages + additional_instructions
-        response: ChatResponse = chat(
-            model=self.client,
-            messages=[
-                {
-                    "role": "user",
-                    "content": message,
-                },
-            ],
-        )
-        return response.message.content
-
-    def generate_audio(self, prompt, **kwargs):
-        raise NotImplementedError
-
-    def generate_image(self, prompt, **kwargs):
-        raise NotImplementedError
-
-    def summarize_text(self, text, primer=""):
-        response: ChatResponse = chat(
-            model=self.client,
-            messages=[
-                {
-                    "role": "system",
-                    "content": f"You are a highly skilled AI trained in language comprehension and summarization.{primer}",
-                },
-                {"role": "user", "content": text},
-            ],
-        )
-        return response.message.content