user_prpmpt at end

Author: gautam-e

README.md

@@ -58,32 +58,29 @@ from pydantic import BaseModel
 ```
 
 ``` python
-model="azure/gpt-4o-2024-08-06"
+model="azure/gpt-4o-2024-08-06" # e.g. openai/gpt-4o-2024-08-06 would use the standard OpenAI
 system_prompt = "Extract the event information."
-```
 
-``` python
-# This defines the output response format that we want
 class CalendarEvent(BaseModel):
     name: str
     date: str
     participants: list[str]
 ```
 
 ``` python
-user_prompt = "Alice and Bob are going to Carmen's Birtday party on 22nd March 2025"
+user_prompt = "Alice and Bob are going to Carmen's birthday party on 22nd March 2025"
 ```
 
 ``` python
 r = structured_output(model=model,
                       system_prompt=system_prompt,
-                      user_prompt=user_prompt,
                       response_format=CalendarEvent, #Note this is the class name (without the `()`)
+                      user_prompt=user_prompt,
                  )
 
 r.model_dump()
 ```
 
     {'name': "Carmen's Birthday Party",
-     'date': '22nd March 2025',
+     'date': '2025-03-22',
      'participants': ['Alice', 'Bob']}

litestruct/__init__.py

@@ -1,2 +1,2 @@
-__version__ = "0.0.2"
+__version__ = "0.0.3"
 from .core import *

litestruct/core.py

@@ -10,20 +10,22 @@
 from litellm import completion
 from pydantic import BaseModel
 
-# %% ../nbs/00_core.ipynb 4
-def structured_output(model:str, # Model name, see examples here or LiteLLM docs for complete list
-                      system_prompt:str, # Instructions for LLM to process the input string
-                      user_prompt:str, # Input string that will be processed
-                      response_format:BaseModel # User-defined Pydantic model to define output 
-                     ) -> BaseModel :
+# %% ../nbs/00_core.ipynb 5
+def structured_output(
+    model:str, # Model name, see examples here or LiteLLM docs for complete list
+    system_prompt:str, # Instructions for LLM to process the input string
+    response_format:BaseModel, # User-defined Pydantic model to define output 
+    user_prompt:str, # Input string that will be processed
+) -> BaseModel:
+    
+    """Get structured output from `model` by combining system and user prompts and making the right API call.
+    See, [here](https://docs.litellm.ai/docs/completion/json_mode#pass-in-json_schema) for full list of APIs available."""
+    
     response = completion(
         model=model,
         messages=[
             {"role": "system", "content": system_prompt},
-            {
-                "role": "user",
-                "content": user_prompt
-            }
+            {"role": "user", "content": user_prompt}
         ],
         response_format=response_format
     )

nbs/00_core.ipynb

@@ -24,8 +24,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "#| hide\n",
-    "from nbdev.showdoc import *"
+    "# | hide\n",
+    "%load_ext autoreload\n",
+    "%autoreload 2"
    ]
   },
   {
@@ -35,45 +36,56 @@
    "outputs": [],
    "source": [
     "#| export\n",
+    "\n",
     "import os\n",
     "from litellm import completion\n",
     "from pydantic import BaseModel"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#| hide\n",
+    "from nbdev.showdoc import *"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "#| export\n",
-    "def structured_output(model:str, # Model name, see examples here or LiteLLM docs for complete list\n",
-    "                      system_prompt:str, # Instructions for LLM to process the input string\n",
-    "                      user_prompt:str, # Input string that will be processed\n",
-    "                      response_format:BaseModel # User-defined Pydantic model to define output \n",
-    "                     ) -> BaseModel :\n",
+    "\n",
+    "def structured_output(\n",
+    "    model:str, # Model name, see examples here or LiteLLM docs for complete list\n",
+    "    system_prompt:str, # Instructions for LLM to process the input string\n",
+    "    response_format:BaseModel, # User-defined Pydantic model to define output \n",
+    "    user_prompt:str, # Input string that will be processed\n",
+    ") -> BaseModel:\n",
+    "    \n",
+    "    \"\"\"Get structured output from `model` by combining system and user prompts and making the right API call.\n",
+    "    See, [here](https://docs.litellm.ai/docs/completion/json_mode#pass-in-json_schema) for full list of APIs available.\"\"\"\n",
+    "    \n",
     "    response = completion(\n",
     "        model=model,\n",
     "        messages=[\n",
     "            {\"role\": \"system\", \"content\": system_prompt},\n",
-    "            {\n",
-    "                \"role\": \"user\",\n",
-    "                \"content\": user_prompt\n",
-    "            }\n",
+    "            {\"role\": \"user\", \"content\": user_prompt}\n",
     "        ],\n",
     "        response_format=response_format\n",
     "    )\n",
     "    return response_format.model_validate_json(response.choices[0].message.content)"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "model=\"azure/gpt-4o-2024-08-06\"\n",
-    "system_prompt = \"Extract the event information.\""
+    "You can control the model that is used for the call by simply adjusting the string in the `model` variable."
    ]
   },
   {
@@ -82,7 +94,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# This defines the output response format that we want\n",
+    "model=\"azure/gpt-4o-2024-08-06\" # e.g. openai/gpt-4o-2024-08-06 would use the standard OpenAI\n",
+    "system_prompt = \"Extract the event information.\"\n",
+    "\n",
     "class CalendarEvent(BaseModel):\n",
     "    name: str\n",
     "    date: str\n",
@@ -95,8 +109,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "user_prompt = \"Alice and Bob are going to Carmen's Birtday party on 22nd March 2025\"\n",
-    "# Note that the word \"Birthday\" is spelt incorrectly on purpose for this example"
+    "user_prompt = \"Alice and Bob are going to Carmen's birthday party on 22nd March 2025\""
    ]
   },
   {
@@ -108,7 +121,7 @@
      "data": {
       "text/plain": [
        "{'name': \"Carmen's Birthday Party\",\n",
-       " 'date': '22nd March 2025',\n",
+       " 'date': '2025-03-22',\n",
        " 'participants': ['Alice', 'Bob']}"
       ]
      },
@@ -121,8 +134,8 @@
     "#| eval: false\n",
     "r = structured_output(model=model,\n",
     "                      system_prompt=system_prompt,\n",
-    "                      user_prompt=user_prompt,\n",
     "                      response_format=CalendarEvent, #Note this is the class name (without the `()`)\n",
+    "                      user_prompt=user_prompt,\n",
     "                 )\n",
     "\n",
     "r.model_dump()"

nbs/index.ipynb

@@ -142,17 +142,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "model=\"azure/gpt-4o-2024-08-06\"\n",
-    "system_prompt = \"Extract the event information.\""
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# This defines the output response format that we want\n",
+    "model=\"azure/gpt-4o-2024-08-06\" # e.g. openai/gpt-4o-2024-08-06 would use the standard OpenAI\n",
+    "system_prompt = \"Extract the event information.\"\n",
+    "\n",
     "class CalendarEvent(BaseModel):\n",
     "    name: str\n",
     "    date: str\n",
@@ -165,7 +157,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "user_prompt = \"Alice and Bob are going to Carmen's Birtday party on 22nd March 2025\""
+    "user_prompt = \"Alice and Bob are going to Carmen's birthday party on 22nd March 2025\""
    ]
   },
   {
@@ -177,7 +169,7 @@
      "data": {
       "text/plain": [
        "{'name': \"Carmen's Birthday Party\",\n",
-       " 'date': '22nd March 2025',\n",
+       " 'date': '2025-03-22',\n",
        " 'participants': ['Alice', 'Bob']}"
       ]
      },
@@ -190,12 +182,19 @@
     "#| eval: false\n",
     "r = structured_output(model=model,\n",
     "                      system_prompt=system_prompt,\n",
-    "                      user_prompt=user_prompt,\n",
     "                      response_format=CalendarEvent, #Note this is the class name (without the `()`)\n",
+    "                      user_prompt=user_prompt,\n",
     "                 )\n",
     "\n",
     "r.model_dump()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

settings.ini

@@ -1,7 +1,7 @@
 [DEFAULT]
 repo = litestruct
 lib_name = litestruct
-version = 0.0.2
+version = 0.0.3
 min_python = 3.9
 license = apache2
 black_formatting = False