New Experimental Training Features, Providers Refactor (#1155)

* Lean AGiXT * Leannnnnnnn * Add defaults * Add to default embedder * Remove requirement * Clean up * Add provider services * Add service endpoint * Add services to agent * Add to defaults * Add provider settings * Fix bad refs * Add services test * Switch model to gpt-4-vision-preview if images uploaded * Remove old refs * Add image generations endpoint * QA Improvements for DPO training * Update dataset creation to use DPO method * Remove embeddings, clean up * Move functionality * Add np * Attempt to fix embedder * Remove old voice code * Maybe fix embedder * Use input instead of text * Tuning wip * Add fine_tune_llm function * Attempt at fixing embedders * Use memories directly * Add auto installs * Clean up * Clean up * Added huggingface api key from agent settings * Update embedding provider * Use messages style for good/bad answers * Add .embeddings * Add static methods for services * Lint * Add training status * Remove text * Fix ref * this seems crazy * Improve training endpoint * v1.5.0 * Increase sleep on test * Fix api_base ref * Clean up provider docs * Fix warning about model_name * Update agixtsdk to 0.0.34 * Update requirements * Add claude vision * Add system message * Clean up readme * Improve dev compose * Add dev pg * Remove code causing error * Updates * Handle provider setting update * Fix get_agents * Add status for user agents * Modify get_agents * Fix duplicates in agent list * Remove condition causing issue * Add tests, fix endpoints * Improve tests * Fix translate_audio response * Fix transcribe audio * Add completions tests to tests on main * Trnascription fix attempt * Convert audio * Fix endpoints * work? * Switch x-wav to wav * Updates * Use file extension * Use wav * Use full path * Always segment * Add dot * fix tts return * Fix tts response * add decoder * STT Fix attempt * Use ffmpeg * Fix test * Clean up * wOrKs oN mY mAcHiNe * Add providers to example * use if not * use agent_name * Use if not user * Add user to collection name * get default user global * Updates * change file type * pass thru urls * Updates * Clean up ezlocalai response * improve vision * change output location for file in test * Try using gtts as default? * Use audiosegment * Fix google provider * Fix api key for google provider * continue * Fix defaults * Add mode * Hope this works! * Fix tabbing * Remove unused refs
Josh-XT · Apr 2, 2024 · 409909b · 409909b
1 parent a16b197
commit 409909b
Show file tree

Hide file tree

Showing 68 changed files with 1,912 additions and 3,264 deletions.
diff --git a/.github/workflows/publish-docker-dev.yml b/.github/workflows/publish-docker-dev.yml
@@ -66,4 +66,13 @@ jobs:
       report-name: "agixt-db-tests"
       additional-python-dependencies: agixtsdk
     needs: build-agixt
-
+  test-completions:
+      uses: josh-xt/AGiXT/.github/workflows/operation-test-with-jupyter.yml@main
+      with:
+        notebook: tests/completions-tests.ipynb
+        image: ghcr.io/${{ needs.build-agixt.outputs.github_user }}/${{ needs.build-agixt.outputs.repo_name }}:${{ github.sha }}
+        port: "7437"
+        db-connected: false
+        report-name: "completions-tests"
+        additional-python-dependencies: openai requests python-dotenv
+      needs: build-agixt
diff --git a/.github/workflows/publish-docker.yml b/.github/workflows/publish-docker.yml
@@ -46,4 +46,13 @@ jobs:
       report-name: "agixt-db-tests"
       additional-python-dependencies: agixtsdk
     needs: build-agixt
-
+  test-completions:
+    uses: josh-xt/AGiXT/.github/workflows/operation-test-with-jupyter.yml@main
+    with:
+      notebook: tests/completions-tests.ipynb
+      image: ${{ needs.build-agixt.outputs.primary-image }}
+      port: "7437"
+      db-connected: false
+      report-name: "completions-tests"
+      additional-python-dependencies: openai requests python-dotenv
+    needs: build-agixt
diff --git a/agixt/AudioToText.py b/agixt/AudioToText.py
diff --git a/agixt/Defaults.py b/agixt/Defaults.py
@@ -6,16 +6,23 @@
 
 DEFAULT_SETTINGS = {
     "provider": "gpt4free",
-    "embedder": "default",
-    "AI_MODEL": "gpt-3.5-turbo",
+    "mode": "prompt",
+    "prompt_category": "Default",
+    "prompt_name": "Chat",
+    "embeddings_provider": "default",
+    "tts_provider": "default",
+    "transcription_provider": "default",
+    "translation_provider": "default",
+    "image_provider": "default",
+    "VOICE": "Brian",
+    "AI_MODEL": "mixtral-8x7b",
     "AI_TEMPERATURE": "0.7",
     "AI_TOP_P": "1",
     "MAX_TOKENS": "4096",
     "helper_agent_name": "gpt4free",
     "WEBSEARCH_TIMEOUT": 0,
     "WAIT_BETWEEN_REQUESTS": 1,
     "WAIT_AFTER_FAILURE": 3,
-    "stream": False,
     "WORKING_DIRECTORY": "./WORKSPACE",
     "WORKING_DIRECTORY_RESTRICTED": True,
     "AUTONOMOUS_EXECUTION": True,

diff --git a/agixt/Embedding.py b/agixt/Embedding.py
diff --git a/agixt/Extensions.py b/agixt/Extensions.py
@@ -145,25 +145,6 @@ def get_commands_list(self):
         return commands_list
 
     async def execute_command(self, command_name: str, command_args: dict = None):
-        if command_args:
-            if "is_m4a_audio" in command_args or "is_wav_audio" in command_args:
-                new_command_args = {}
-                for arg in command_args:
-                    if "is_m4a_audio" in command_args:
-                        if arg == command_args["is_m4a_audio"]:
-                            new_command_args[arg] = await self.execute_command(
-                                command_name="Transcribe M4A Audio",
-                                command_args={"base64_audio": command_args[arg]},
-                            )
-                    if "is_wav_audio" in command_args:
-                        if arg == command_args["is_wav_audio"]:
-                            new_command_args[arg] = await self.execute_command(
-                                command_name="Transcribe WAV Audio",
-                                command_args={"base64_audio": command_args[arg]},
-                            )
-                    if arg not in new_command_args:
-                        new_command_args[arg] = command_args[arg]
-                command_args = new_command_args
         injection_variables = {
             "user": self.user,
             "agent_name": self.agent_name,

diff --git a/agixt/Interactions.py b/agixt/Interactions.py
@@ -409,27 +409,9 @@ async def run(
         browse_links: bool = False,
         prompt_category: str = "Default",
         persist_context_in_history: bool = False,
-        is_m4a_audio: bool = False,
-        is_wav_audio: bool = False,
         images: list = [],
         **kwargs,
     ):
-        if is_m4a_audio or is_wav_audio:
-            # Convert the audio to text
-            ext = Extensions(
-                agent_name=self.agent_name,
-                agent_config=self.agent.AGENT_CONFIG,
-                conversation_name=conversation_name,
-                ApiClient=self.ApiClient,
-                user=self.user,
-            )
-            command_output = await ext.execute_command(
-                command_name=(
-                    "Transcribe M4A Audio" if is_m4a_audio else "Transcribe WAV Audio"
-                ),
-                command_args={"base64_audio": user_input},
-            )
-            user_input = command_output
         shots = int(shots)
         if "prompt_category" in kwargs:
             prompt_category = kwargs["prompt_category"]