Merge branch 'test/demo' into feat/model

Author: jjmachan

.gitignore

@@ -172,3 +172,4 @@ cython_debug/
 .python-version
 uv.lock
 _proc
+experiments

nbs/backends/ragas_api_client.ipynb

@@ -1069,7 +1069,7 @@
        " 'datatable_id': 'ebc3dd3e-f88b-4f8b-8c72-6cfcae0a0cd4'}"
       ]
      },
-     "execution_count": 32,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -1379,6 +1379,63 @@
    "execution_count": 40,
    "metadata": {},
    "outputs": [],
+   "source": [
+    "#| export\n",
+    "import uuid\n",
+    "import string"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#| export\n",
+    "def create_nano_id(size=12):\n",
+    "    # Define characters to use (alphanumeric)\n",
+    "    alphabet = string.ascii_letters + string.digits\n",
+    "    \n",
+    "    # Generate UUID and convert to int\n",
+    "    uuid_int = uuid.uuid4().int\n",
+    "    \n",
+    "    # Convert to base62\n",
+    "    result = \"\"\n",
+    "    while uuid_int:\n",
+    "        uuid_int, remainder = divmod(uuid_int, len(alphabet))\n",
+    "        result = alphabet[remainder] + result\n",
+    "    \n",
+    "    # Pad if necessary and return desired length\n",
+    "    return result[:size]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'Anvz5k9geU7T'"
+      ]
+     },
+     "execution_count": null,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# Usage\n",
+    "nano_id = create_nano_id()  # e.g., \"8dK9cNw3mP5x\"\n",
+    "nano_id"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "#| export\n",
     "# Default settings for columns\n",

nbs/metric/base.ipynb

@@ -42,8 +42,9 @@
     "from dataclasses import dataclass, field\n",
     "from pydantic import BaseModel\n",
     "import typing as t\n",
-    "import json\n",
     "from tqdm import tqdm\n",
+    "import string\n",
+    "\n",
     "\n",
     "from ragas_annotator.prompt.base import Prompt\n",
     "from ragas_annotator.embedding.base import BaseEmbedding\n",
@@ -76,7 +77,14 @@
     "    @abstractmethod\n",
     "    def _ensemble(self, results: t.List[MetricResult]) -> MetricResult:\n",
     "        pass\n",
-    "        \n",
+    "    \n",
+    "    def get_variables(self) -> t.List[str]:\n",
+    "        if isinstance(self.prompt, Prompt):\n",
+    "            fstr = self.prompt.instruction\n",
+    "        else:\n",
+    "            fstr = self.prompt\n",
+    "        vars = [field_name for _, field_name, _, _ in string.Formatter().parse(fstr) if field_name]\n",
+    "        return vars\n",
     "    \n",
     "    def score(self, reasoning: bool = True, n: int = 1, **kwargs) -> t.Any:\n",
     "        responses = []\n",
@@ -130,13 +138,15 @@
     "            datasets.append(experiment_data)\n",
     "        \n",
     "        total_items = sum([len(dataset) for dataset in datasets])\n",
+    "        input_vars = self.get_variables()\n",
+    "        output_vars = [self.name, f'{self.name}_reason']\n",
     "        with tqdm(total=total_items, desc=\"Processing examples\") as pbar:\n",
     "            for dataset in datasets:\n",
     "                for row in dataset:\n",
-    "                    if hasattr(row, f'{self.name}_traces'):\n",
-    "                        traces = json.loads(getattr(row, f'{self.name}_traces'))\n",
-    "                        if traces:\n",
-    "                            self.prompt.add_example(traces['input'],traces['output'])\n",
+    "                    inputs = {var: getattr(row, var) for var in input_vars if hasattr(row, var)}\n",
+    "                    output = {var: getattr(row, var) for var in output_vars if hasattr(row, var)}\n",
+    "                    if output:\n",
+    "                            self.prompt.add_example(inputs,output)\n",
     "                    pbar.update(1)\n",
     "        \n",
     "                \n",
@@ -160,7 +170,18 @@
    "execution_count": null,
    "id": "fcf208fa",
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "100"
+      ]
+     },
+     "execution_count": null,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
    "source": [
     "#| eval: false\n",
     "\n",
@@ -189,6 +210,13 @@
     "my_metric = CustomMetric(name=\"example\", prompt=\"What is the result of {input}?\", llm=llm)\n",
     "my_metric.score(input=\"test\")"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

ragas_annotator/_modidx.py

@@ -297,6 +297,8 @@
                                                                                             'ragas_annotator/metric/base.py'),
                                              'ragas_annotator.metric.base.Metric.batch_score': ( 'metric/base.html#metric.batch_score',
                                                                                                  'ragas_annotator/metric/base.py'),
+                                             'ragas_annotator.metric.base.Metric.get_variables': ( 'metric/base.html#metric.get_variables',
+                                                                                                   'ragas_annotator/metric/base.py'),
                                              'ragas_annotator.metric.base.Metric.score': ( 'metric/base.html#metric.score',
                                                                                            'ragas_annotator/metric/base.py'),
                                              'ragas_annotator.metric.base.Metric.train': ( 'metric/base.html#metric.train',

ragas_annotator/metric/base.py

@@ -11,8 +11,9 @@
 from dataclasses import dataclass, field
 from pydantic import BaseModel
 import typing as t
-import json
 from tqdm import tqdm
+import string
+
 
 from ..prompt.base import Prompt
 from ..embedding.base import BaseEmbedding
@@ -45,7 +46,14 @@ def _get_response_model(self, with_reasoning: bool) -> t.Type[BaseModel]:
     @abstractmethod
     def _ensemble(self, results: t.List[MetricResult]) -> MetricResult:
         pass
-        
+    
+    def get_variables(self) -> t.List[str]:
+        if isinstance(self.prompt, Prompt):
+            fstr = self.prompt.instruction
+        else:
+            fstr = self.prompt
+        vars = [field_name for _, field_name, _, _ in string.Formatter().parse(fstr) if field_name]
+        return vars
 
     def score(self, reasoning: bool = True, n: int = 1, **kwargs) -> t.Any:
         responses = []
@@ -99,13 +107,15 @@ def train(self,project:Project, experiment_names: t.List[str], model:NotionModel
             datasets.append(experiment_data)
 
         total_items = sum([len(dataset) for dataset in datasets])
+        input_vars = self.get_variables()
+        output_vars = [self.name, f'{self.name}_reason']
         with tqdm(total=total_items, desc="Processing examples") as pbar:
             for dataset in datasets:
                 for row in dataset:
-                    if hasattr(row, f'{self.name}_traces'):
-                        traces = json.loads(getattr(row, f'{self.name}_traces'))
-                        if traces:
-                            self.prompt.add_example(traces['input'],traces['output'])
+                    inputs = {var: getattr(row, var) for var in input_vars if hasattr(row, var)}
+                    output = {var: getattr(row, var) for var in output_vars if hasattr(row, var)}
+                    if output:
+                            self.prompt.add_example(inputs,output)
                     pbar.update(1)