nosible 0.2.1__py3-none-any.whl → 0.2.2__py3-none-any.whl

nosible/classes/result.py

@@ -3,9 +3,8 @@ from __future__ import annotations
 from dataclasses import asdict, dataclass
 from typing import TYPE_CHECKING
 
-from openai import OpenAI
-
 from nosible.classes.web_page import WebPageData
+from nosible.utils.json_tools import print_dict
 
 if TYPE_CHECKING:
     from nosible.classes.result_set import ResultSet
@@ -102,11 +101,21 @@ class Result:
           0.99 | Example Domain
         >>> result = Result(title=None, similarity=None)
         >>> print(str(result))
-           N/A | No Title
+        {
+            "url": null,
+            "title": null,
+            "description": null,
+            "netloc": null,
+            "published": null,
+            "visited": null,
+            "author": null,
+            "content": null,
+            "language": null,
+            "similarity": null,
+            "url_hash": null
+        }
         """
-        similarity = f"{self.similarity:.2f}" if self.similarity is not None else "N/A"
-        title = self.title or "No Title"
-        return f"{similarity:>6} | {title}"
+        return print_dict(self.to_dict())
 
     def __getitem__(self, key: str) -> str | float | bool | None:
         """
@@ -295,12 +304,12 @@ class Result:
 
             The response must be a float in [-1.0, 1.0]. No other text must be returned.
         """
-
+        from openai import OpenAI
         llm_client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=client.llm_api_key)
 
         # Call the chat completions endpoint.
         resp = llm_client.chat.completions.create(
-            model="openai/gpt-4o", messages=[{"role": "user", "content": prompt.strip()}], temperature=0.7
+            model=client.sentiment_model, messages=[{"role": "user", "content": prompt.strip()}], temperature=0.7
         )
 
         raw = resp.choices[0].message.content

nosible/classes/result_set.py

@@ -2,15 +2,15 @@ from __future__ import annotations
 
 from collections.abc import Iterator
 from dataclasses import dataclass, field
-
-import duckdb
-import pandas as pd
-import polars as pl
-from tantivy import Document, Index, SchemaBuilder
+from typing import TYPE_CHECKING
 
 from nosible.classes.result import Result
 from nosible.utils.json_tools import json_dumps, json_loads
 
+if TYPE_CHECKING:
+    import pandas as pd
+    import polars as pl
+
 
 @dataclass(frozen=True)
 class ResultSet(Iterator[Result]):
@@ -182,28 +182,34 @@ class ResultSet(Iterator[Result]):
         # Setup if required
         return self
 
-    def __getitem__(self, key: int) -> Result:
+    def __getitem__(self, key: int | slice) -> Result | ResultSet:
         """
-        Get a Result by index.
+        Get a Result by index or a list of Results by slice.
 
         Parameters
         ----------
-        key : int
-            Index of the result to retrieve.
+        key : int or slice
+            Index or slice of the result(s) to retrieve.
 
         Returns
         -------
-        Result
-            The Result at the specified index.
+        Result or ResultSet
+            A single Result if `key` is an integer, or a ResultSet containing the sliced results if `key` is a slice.
 
         Raises
         ------
         IndexError
             If index is out of range.
+        TypeError
+            If key is not an integer or slice.
         """
-        if 0 <= key < len(self.results):
-            return self.results[key]
-        raise IndexError(f"Index {key} out of range for ResultSet with length {len(self.results)}.")
+        if isinstance(key, int):
+            if 0 <= key < len(self.results):
+                return self.results[key]
+            raise IndexError(f"Index {key} out of range for ResultSet with length {len(self.results)}.")
+        if isinstance(key, slice):
+            return ResultSet(self.results[key])
+        raise TypeError("ResultSet indices must be integers or slices.")
 
     def __add__(self, other: ResultSet | Result) -> ResultSet:
         """
@@ -316,6 +322,8 @@ class ResultSet(Iterator[Result]):
         Document returned
         Document returned
         """
+        from tantivy import Document, Index, SchemaBuilder
+
         # Build the Tantivy schema
         schema_builder = SchemaBuilder()
         # Int for doc retrieval.
@@ -439,6 +447,9 @@ class ResultSet(Iterator[Result]):
         Traceback (most recent call last):
         ValueError: Cannot analyze by 'foobar' - not a valid field.
         """
+        import pandas as pd
+        import polars as pl
+
         # Convert to Polars DataFrame
         df: pl.DataFrame = self.to_polars()
 
@@ -571,6 +582,10 @@ class ResultSet(Iterator[Result]):
         >>> "url" in df.columns
         True
         """
+        # Lazy import for runtime, but allow static type checking
+
+        import polars as pl
+
         return pl.DataFrame(self.to_dicts())
 
     def to_pandas(self) -> pd.DataFrame:
@@ -911,7 +926,7 @@ class ResultSet(Iterator[Result]):
             import duckdb
 
             # Convert to Polars DataFrame and then to Arrow Table
-            df = self.to_polars()
+            df = self.to_polars()  # noqa: F841
             # Connect to DuckDB and write the Arrow Table to a table
             con = duckdb.connect(out)
             # Write the DataFrame to the specified table name, replacing if exists
@@ -964,6 +979,8 @@ class ResultSet(Iterator[Result]):
         >>> results[0].title
         'Example Domain'
         """
+        import polars as pl
+
         try:
             df = pl.read_csv(file_path)
         except Exception as e:
@@ -1124,6 +1141,8 @@ class ResultSet(Iterator[Result]):
         >>> print(len(df))
         1
         """
+        import polars as pl
+
         pl_df = pl.from_pandas(df)
         return cls.from_polars(pl_df)
 
@@ -1239,6 +1258,8 @@ class ResultSet(Iterator[Result]):
         >>> results[0].title
         'Example Domain'
         """
+        import polars as pl
+
         try:
             df = pl.read_parquet(file_path)
         except Exception as e:
@@ -1288,6 +1309,8 @@ class ResultSet(Iterator[Result]):
         >>> results[0].title
         'Example Domain'
         """
+        import polars as pl
+
         try:
             df = pl.read_ipc(file_path)
         except Exception as e:
@@ -1340,7 +1363,11 @@ class ResultSet(Iterator[Result]):
         >>> loaded[0].title
         'Example Domain'
         """
+        import polars as pl
+
         try:
+            import duckdb
+
             con = duckdb.connect(file_path, read_only=True)
         except Exception as e:
             raise RuntimeError(f"Failed to connect to DuckDB file '{file_path}': {e}") from e
@@ -1492,10 +1519,3 @@ class ResultSet(Iterator[Result]):
         """
         # TODO: cleanup handles, sessions, etc.
         pass
-
-
-if __name__ == "__main__":
-    import doctest
-
-    doctest.testmod(optionflags=doctest.ELLIPSIS | doctest.NORMALIZE_WHITESPACE)
-    print("All tests passed!")

nosible/nosible_client.py

@@ -2,21 +2,17 @@ import gzip
 import json
 import logging
 import os
+import re
 import sys
 import textwrap
 import time
 import types
-import typing
 from collections.abc import Iterator
 from concurrent.futures import ThreadPoolExecutor
 from datetime import datetime
-from typing import Union, Optional
+from typing import Optional, Union
 
-import polars as pl
-import requests
-from cryptography.fernet import Fernet
-from openai import OpenAI
-from polars import SQLContext
+import httpx
 from tenacity import (
     before_sleep_log,
     retry,
@@ -32,7 +28,6 @@ from nosible.classes.search_set import SearchSet
 from nosible.classes.snippet_set import SnippetSet
 from nosible.classes.web_page import WebPageData
 from nosible.utils.json_tools import json_loads
-from nosible.utils.question_builder import _get_question
 from nosible.utils.rate_limiter import PLAN_RATE_LIMITS, RateLimiter, _rate_limited
 
 # Set up a module‐level logger.
@@ -56,6 +51,8 @@ class Nosible:
         Base URL for the OpenAI-compatible LLM API. (default is OpenRouter's API endpoint)
     sentiment_model : str, optional
         Model to use for sentiment analysis (default is "openai/gpt-4o").
+    expansions_model : str, optional
+        Model to use for expansions (default is "openai/gpt-4o").
     timeout : int
         Request timeout for HTTP calls.
     retries : int,
@@ -94,7 +91,8 @@ class Nosible:
     - The `nosible_api_key` is required to access the Nosible Search API.
     - The `llm_api_key` is optional and used for LLM-based query expansions.
     - The `openai_base_url` defaults to OpenRouter's API endpoint.
-    - The `sentiment_model` is used for generating query expansions and sentiment analysis.
+    - The `sentiment_model` is used for sentiment analysis.
+    - The `expansions_model` is used for generating query expansions.
     - The `timeout`, `retries`, and `concurrency` parameters control the behavior of HTTP requests.
 
     Examples
@@ -106,10 +104,11 @@ class Nosible:
 
     def __init__(
         self,
-        nosible_api_key: str = None,
-        llm_api_key: str = None,
+        nosible_api_key: Optional[str] = None,
+        llm_api_key: Optional[str] = None,
         openai_base_url: str = "https://openrouter.ai/api/v1",
         sentiment_model: str = "openai/gpt-4o",
+        expansions_model: str = "openai/gpt-4o",
         timeout: int = 30,
         retries: int = 5,
         concurrency: int = 10,
@@ -142,6 +141,7 @@ class Nosible:
         self.llm_api_key = llm_api_key or os.getenv("LLM_API_KEY")
         self.openai_base_url = openai_base_url
         self.sentiment_model = sentiment_model
+        self.expansions_model = expansions_model
         # Network parameters
         self.timeout = timeout
         self.retries = retries
@@ -162,7 +162,7 @@ class Nosible:
             reraise=True,
             stop=stop_after_attempt(self.retries) | stop_after_delay(self.timeout),
             wait=wait_exponential(multiplier=1, min=1, max=10),
-            retry=retry_if_exception_type(requests.exceptions.RequestException),
+            retry=retry_if_exception_type(httpx.RequestError),
             before_sleep=before_sleep_log(self.logger, logging.WARNING),
         )(self._post)
 
@@ -171,12 +171,12 @@ class Nosible:
             reraise=True,
             stop=stop_after_attempt(self.retries) | stop_after_delay(self.timeout),
             wait=wait_exponential(multiplier=1, min=1, max=10),
-            retry=retry_if_exception_type(requests.exceptions.RequestException),
+            retry=retry_if_exception_type(httpx.RequestError),
             before_sleep=before_sleep_log(self.logger, logging.WARNING),
         )(self._generate_expansions)
 
         # Thread pool for parallel searches
-        self._session = requests.Session()
+        self._session = httpx.Client(follow_redirects=True)
         self._executor = ThreadPoolExecutor(max_workers=self.concurrency)
 
         # Headers
@@ -201,7 +201,6 @@ class Nosible:
 
     def search(
         self,
-        *,
         search: Search = None,
         question: str = None,
         expansions: list[str] = None,
@@ -873,6 +872,8 @@ class Nosible:
         ...
         ValueError: Bulk search cannot have more than 10000 results per query.
         """
+        from cryptography.fernet import Fernet
+
         previous_level = self.logger.level
         if verbose:
             self.logger.setLevel(logging.INFO)
@@ -981,7 +982,7 @@ class Nosible:
             resp = self._post(url="https://www.nosible.ai/search/v1/slow-search", payload=payload)
             try:
                 resp.raise_for_status()
-            except requests.HTTPError as e:
+            except httpx.HTTPStatusError as e:
                 raise ValueError(f"[{question!r}] HTTP {resp.status_code}: {resp.text}") from e
 
             data = resp.json()
@@ -993,7 +994,7 @@ class Nosible:
             decrypt_using = data.get("decrypt_using")
             for _ in range(100):
                 dl = self._session.get(download_from, timeout=self.timeout)
-                if dl.ok:
+                if dl.status_code == 200:
                     fernet = Fernet(decrypt_using.encode())
                     decrypted = fernet.decrypt(dl.content)
                     decompressed = gzip.decompress(decrypted)
@@ -1053,7 +1054,7 @@ class Nosible:
         ...     ans = nos.answer(
         ...         query="How is research governance and decision-making structured between Google and DeepMind?",
         ...         n_results=100,
-        ...         show_context=True
+        ...         show_context=True,
         ...     )  # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
         <BLANKLINE>
         Doc 1
@@ -1067,11 +1068,7 @@ class Nosible:
             raise ValueError("An LLM API key is required for answer().")
 
         # Retrieve top documents
-        results = self.search(
-            question=query,
-            n_results=n_results,
-            min_similarity=min_similarity,
-        )
+        results = self.search(question=query, n_results=n_results, min_similarity=min_similarity)
 
         # Build RAG context
         context = ""
@@ -1090,7 +1087,7 @@ class Nosible:
             print(textwrap.dedent(context))
 
         # Craft prompt
-        prompt = (f"""
+        prompt = f"""
             # TASK DESCRIPTION
 
             You are a helpful assistant.  Use the following context to answer the question.
@@ -1102,15 +1099,12 @@ class Nosible:
             ## Context
             {context}
             """
-        )
+        from openai import OpenAI
 
         # Call LLM
         client = OpenAI(base_url=self.openai_base_url, api_key=self.llm_api_key)
         try:
-            response = client.chat.completions.create(
-                model = model,
-                messages = [{"role": "user", "content": prompt}],
-            )
+            response = client.chat.completions.create(model=model, messages=[{"role": "user", "content": prompt}])
         except Exception as e:
             raise RuntimeError(f"LLM API error: {e}") from e
 
@@ -1123,13 +1117,7 @@ class Nosible:
         return "Answer:\n" + response.choices[0].message.content.strip()
 
     @_rate_limited("visit")
-    def visit(
-        self,
-        html: str = "",
-        recrawl: bool = False,
-        render: bool = False,
-        url: str = None
-    ) -> WebPageData:
+    def visit(self, html: str = "", recrawl: bool = False, render: bool = False, url: str = None) -> WebPageData:
         """
         Visit a given URL and return a structured WebPageData object for the page.
 
@@ -1262,10 +1250,7 @@ class Nosible:
             payload["sql_filter"] = "SELECT loc, published FROM engine"
 
         # Send the POST to the /trend endpoint
-        response = self._post(
-            url="https://www.nosible.ai/search/v1/trend",
-            payload=payload,
-        )
+        response = self._post(url="https://www.nosible.ai/search/v1/trend", payload=payload)
         # Will raise ValueError on rate-limit or auth errors
         response.raise_for_status()
         payload = response.json().get("response", {})
@@ -1365,7 +1350,7 @@ class Nosible:
                 return False
             # If we reach here, the response is unexpected
             return False
-        except requests.HTTPError:
+        except httpx.HTTPError:
             return False
         except:
             return False
@@ -1460,7 +1445,7 @@ class Nosible:
         out = [
             "Below are the rate limits for all NOSIBLE plans.",
             "To upgrade your package, visit https://www.nosible.ai/products.\n",
-            "Unless otherwise indicated, bulk searches are limited to one-at-a-time per API key.\n"
+            "Unless otherwise indicated, bulk searches are limited to one-at-a-time per API key.\n",
         ]
 
         user_plan = self._get_user_plan()
@@ -1521,7 +1506,7 @@ class Nosible:
         except Exception:
             pass
 
-    def _post(self, url: str, payload: dict, headers: dict = None, timeout: int = None) -> requests.Response:
+    def _post(self, url: str, payload: dict, headers: dict = None, timeout: int = None) -> httpx.Response:
         """
         Internal helper to send a POST request with retry logic.
 
@@ -1553,7 +1538,7 @@ class Nosible:
 
         Returns
         -------
-        requests.Response
+        httpx.Response
             The HTTP response object.
         """
         response = self._session.post(
@@ -1561,18 +1546,18 @@ class Nosible:
             json=payload,
             headers=headers if headers is not None else self.headers,
             timeout=timeout if timeout is not None else self.timeout,
+            follow_redirects=True,
         )
 
         # If unauthorized, or if the payload is string too short, treat as invalid API key
         if response.status_code == 401:
             raise ValueError("Your API key is not valid.")
         if response.status_code == 422:
-            # Only inspect JSON if it’s a JSON response
             content_type = response.headers.get("Content-Type", "")
             if content_type.startswith("application/json"):
                 body = response.json()
                 if isinstance(body, list):
-                    body = body[0]  # NOSIBLE returns a list of errors
+                    body = body[0]
                 print(body)
                 if body.get("type") == "string_too_short":
                     raise ValueError("Your API key is not valid: Too Short.")
@@ -1711,12 +1696,14 @@ class Nosible:
                - Contextual Example: Swap "diabetes treatment" with "insulin therapy" or "blood sugar management".
 
         """.replace("                ", "")
+        # Lazy load
+        from openai import OpenAI
 
         client = OpenAI(base_url=self.openai_base_url, api_key=self.llm_api_key)
 
         # Call the chat completions endpoint.
         resp = client.chat.completions.create(
-            model=self.sentiment_model, messages=[{"role": "user", "content": prompt.strip()}], temperature=0.7
+            model=self.expansions_model, messages=[{"role": "user", "content": prompt.strip()}], temperature=0.7
         )
 
         raw = resp.choices[0].message.content
@@ -1776,14 +1763,16 @@ class Nosible:
             ...
         ValueError: Invalid date for 'visited_start': '2023/12/31'.  Expected ISO format 'YYYY-MM-DD'.
         """
+        dateregex = r"^\d{4}-\d{2}-\d{2}"
+
+        if not re.match(dateregex, string):
+            raise ValueError(f"Invalid date for '{name}': {string!r}.  Expected ISO format 'YYYY-MM-DD'.")
+
         try:
             # datetime.fromisoformat accepts both YYYY-MM-DD and full timestamps
             parsed = datetime.fromisoformat(string)
         except Exception:
-            raise ValueError(
-                f"Invalid date for '{name}': {string!r}.  "
-                "Expected ISO format 'YYYY-MM-DD'."
-            )
+            raise ValueError(f"Invalid date for '{name}': {string!r}.  Expected ISO format 'YYYY-MM-DD'.")
 
     def _format_sql(
         self,
@@ -1996,9 +1985,11 @@ class Nosible:
             "company_3",
             "doc_hash",
         ]
+        import polars as pl  # Lazy import
+
         # Create a dummy DataFrame with correct columns and no rows
         df = pl.DataFrame({col: [] for col in columns})
-        ctx = SQLContext()
+        ctx = pl.SQLContext()
         ctx.register("engine", df)
         try:
             ctx.execute(sql)
@@ -2019,10 +2010,10 @@ class Nosible:
 
     def __exit__(
         self,
-        _exc_type: typing.Optional[type[BaseException]],
-        _exc_val: typing.Optional[BaseException],
-        _exc_tb: typing.Optional[types.TracebackType],
-    ) -> typing.Optional[bool]:
+        _exc_type: Optional[type[BaseException]],
+        _exc_val: Optional[BaseException],
+        _exc_tb: Optional[types.TracebackType],
+    ) -> Optional[bool]:
         """
         Always clean up (self.close()), but let exceptions propagate.
         Return True only if you really want to suppress an exception.

{nosible-0.2.1.dist-info → nosible-0.2.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nosible
-Version: 0.2.1
+Version: 0.2.2
 Summary: Python client for the NOSIBLE Search API
 Home-page: https://github.com/NosibleAI/nosible
 Author: Stuart Reid, Matthew Dicks, Richard Taylor, Gareth Warburton
@@ -35,8 +35,8 @@ Requires-Dist: tantivy
 Requires-Dist: pyrate-limiter
 Requires-Dist: tenacity
 Requires-Dist: cryptography
-Requires-Dist: pandas
 Requires-Dist: pyarrow
+Requires-Dist: pandas
 Dynamic: author
 Dynamic: home-page
 Dynamic: license-file
@@ -80,13 +80,15 @@ uv pip install nosible
 **Requirements**:
 
 * Python 3.9+
-* requests
 * polars
-* cryptography
-* tenacity
-* pyrate-limiter
-* tantivy
+* duckdb
 * openai
+* tantivy
+* pyrate-limiter
+* tenacity
+* cryptography
+* pyarrow
+* pandas
 
 ### 🔑 Authentication
 
@@ -140,9 +142,28 @@ os.environ["LLM_API_KEY"] = "sk-..."
 
 ### 🚀 Examples
 
-#### Fast Search
+#### Search
+
+The Search and Searches functions enables you to retrieve **up to 100** results for a single query. This is ideal for most use cases where you need to retrieve information quickly and efficiently.
+
+- Use the `search` method when you need between **10 and 100** results for a single query.
+- The same applies for the `searches` and `.similar()` methods.
 
-Retrieve up to 100 results with optional filters:
+- A search will return a set of `Result` objects.
+- The `Result` object is used to represent a single search result and provides methods to access the result's properties.
+    - `url`: The URL of the search result.
+    - `title`: The title of the search result.
+    - `description`: A brief description or summary of the search result.
+    - `netloc`: The network location (domain) of the URL.
+    - `published`: The publication date of the search result.
+    - `visited`: The date and time when the result was visited.
+    - `author`: The author of the content.
+    - `content`: The main content or body of the search result.
+    - `language`: The language code of the content (e.g., 'en' for English).
+    - `similarity`: Similarity score with respect to a query or reference.
+
+They can be accessed directly from the `Result` object: `print(result.title)` or
+`print(result["title"])`
 
 ```python
 from nosible import Nosible
@@ -169,9 +190,44 @@ with Nosible(
     print([r.title for r in results])
 ```
 
+#### Expansions
+
+**Prompt expansions** are questions **lexically** and **semantically similar** to your main question. Expansions are added alongside your search query to improve your search results. You can add up to 10 expansions per search.
+
+- You can add you **own expansions** by passing a list of strings to the `expansions` parameter.
+- You can also get your expansions automatically generated by setting `autogenerate_expansions` to `True` when running the search.
+    - For expansions to be generated, you will need the `LLM_API_KEY` to be set in the environment or passed to the `Nosible` constructor.
+      - By default, we use openrouter as an endpoint. However, **we support any endpoint that supports openai**. If you 
+        want to use a different endpoint, follow [this](https://nosible-py.readthedocs.io/en/latest/configuration.html#change-llm-base-url) guide in the docs.
+    - You can change this model with the argument **expansions_model**.
+
+```python
+# Example of using your own expansions
+with Nosible() as nos:
+    results = nos.search(
+        question="How have the Trump tariffs impacted the US economy?",
+        expansions=[
+            "What are the consequences of Trump's 2018 steel and aluminum tariffs on American manufacturers?",
+            "How did Donald Trump's tariffs on Chinese imports influence US import prices and inflation?",
+            "What impact did the Section 232 tariffs under President Trump have on US agricultural exports?",
+            "In what ways have Trump's trade duties affected employment levels in the US automotive sector?",
+            "How have the tariffs imposed by the Trump administration altered American consumer goods pricing nationwide?",
+            "What economic outcomes resulted from President Trump's protective tariffs for the United States economy?",
+            "How did Trump's solar panel tariffs change investment trends in the US energy market?",
+            "What have been the financial effects of Trump's Section 301 tariffs on Chinese electronics imports?",
+            "How did Trump's trade barriers influence GDP growth and trade deficits in the United States?",
+            "In what manner did Donald Trump's import taxes reshape competitiveness of US steel producers globally?",
+        ],
+        n_results=10,
+    )
+
+print(results)
+```
+
 #### Parallel Searches
 
-Run multiple queries concurrently:
+Allows you to run multiple searches concurrently and `yields` the results as they come in.
+- You can pass a list of questions to the `searches` method.
 
 ```python
 from nosible import Nosible
@@ -190,7 +246,12 @@ with Nosible(nosible_api_key="basic|abcd1234...", llm_api_key="sk-...") as clien
 
 #### Bulk Search
 
-Fetch thousands of results for offline analysis:
+Bulk search enables you to retrieve a large number of results in a single request, making it ideal for large-scale data analysis and processing.
+
+- Use the `bulk_search` method when you need more than 1,000 results for a single query.
+- You can request between **1,000 and 10,000** results per query.
+- All parameters available in the standard `search` method—such as `expansions`, `include_companies`, `include_languages`, and more—are also supported in `bulk_search`.
+- A bulk search for 10,000 results typically completes in about 30 seconds or less.
 
 ```python
 from nosible import Nosible
@@ -244,9 +305,14 @@ with Nosible(nosible_api_key="basic|abcd1234...") as client:
     print([r for r in results])
 ```
 
-#### Sentiment Analysis
+#### Sentiment
 
-Compute sentiment for a single result (uses GPT-4o; requires an LLM API key):
+This fetches a sentiment score for each search result.
+- The sentiment score is a float between `-1` and `1`, where `-1` is **negative**, `0` is **neutral**, and `1` is **positive**.
+- The sentiment model can be changed by passing the `sentiment_model` parameter to the `Nosible` constructor.
+    - The `sentiment_model` defaults to "openai/gpt-4o", which is a powerful model for sentiment analysis.
+- You can also change the base URL for the LLM API by passing the `openai_base_url` parameter to the `Nosible` constructor.
+    - The `openai_base_url` defaults to OpenRouter's API endpoint.
 
 ```python
 from nosible import Nosible

{nosible-0.2.1.dist-info → nosible-0.2.2.dist-info}/RECORD

@@ -1,17 +1,16 @@
 nosible/__init__.py,sha256=11QmG9Wjprp_zB0VnPxGjqKwHmaoB0hoT8AGO6cGVMM,1426
-nosible/nosible_client.py,sha256=IjT0OPmDBF0xlTL4W_njHMo2eYesQu8CBo7cHpuqWE8,82099
-nosible/classes/result.py,sha256=Ao6SLlaPBo6Rhb7WkK59h5C5OGBuz7qhfyHVTtzO1wI,17407
-nosible/classes/result_set.py,sha256=Rr5WBoMot6bYh7AKSZEPCznm0WNfUuPkX1PeHMv8rA4,52160
+nosible/nosible_client.py,sha256=rWwENpe5pXY0v7xiuHzYyK6WeCT4HMV3JDyubUqVhfE,82214
+nosible/classes/result.py,sha256=XNwFAGU_8nCqQNCn-mFRHXzFbUrtTHDsfCyDxjbOTQ8,17652
+nosible/classes/result_set.py,sha256=badc0q8rlvgUQ36ZrNgSY5O40lIvyZsCChtpnoCM5jk,52813
 nosible/classes/search.py,sha256=_RDH8ih75RufztOG7NUJuZ1ncNV3jNThq8ST41VP0fU,11269
 nosible/classes/search_set.py,sha256=4L2qO_IKqlUebvxNJERxjXNiUKtolw7JhgFtt1QqEHk,9790
 nosible/classes/snippet.py,sha256=5Av0cXjOL-8X6H8oFunC36DUyWiFLLsl7FfPJ7cYwVU,4988
 nosible/classes/snippet_set.py,sha256=1SZUEq6zRcBT3N8BKbFPml5pt_WSCybIjMpTw8PTz4E,5093
 nosible/classes/web_page.py,sha256=fe-cfDL-AmQRMdBO37jWh1K6BsH_K_VWBBfBzFRV3KM,7149
 nosible/utils/json_tools.py,sha256=PcSMjcLEhbA626jAIn0SuD_1-4QDduapZUenTSt3N2E,4569
-nosible/utils/question_builder.py,sha256=S3eCb0Dsg0_1kyUmp2hxYnxH4Yq_tUi-HkB8owUB60E,4063
 nosible/utils/rate_limiter.py,sha256=qr0Tg-3wVcw95FyQv3gbZhbf-_QY9zKdkIE4FSLFSBo,5400
-nosible-0.2.1.dist-info/licenses/LICENSE,sha256=8ifsV4DrsiKi8KVBFy8SBb3KXPXhofE3pYq07q1TSCQ,1117
-nosible-0.2.1.dist-info/METADATA,sha256=W2KaKTxdxD6JQSWQgnc_eMtDzTJo0B9CfKBC7AkX7EM,10209
-nosible-0.2.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-nosible-0.2.1.dist-info/top_level.txt,sha256=mOconHuKcNJ1jTAj3DapQP_xB8YOmjTMyHg5txKH3uA,8
-nosible-0.2.1.dist-info/RECORD,,
+nosible-0.2.2.dist-info/licenses/LICENSE,sha256=8ifsV4DrsiKi8KVBFy8SBb3KXPXhofE3pYq07q1TSCQ,1117
+nosible-0.2.2.dist-info/METADATA,sha256=o_cCaHsq8amNEL7HZEKwpu_X_RKFNa06ViqpLo4bsx8,14881
+nosible-0.2.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+nosible-0.2.2.dist-info/top_level.txt,sha256=mOconHuKcNJ1jTAj3DapQP_xB8YOmjTMyHg5txKH3uA,8
+nosible-0.2.2.dist-info/RECORD,,

nosible/utils/question_builder.py

@@ -1,131 +0,0 @@
-import random
-
-COMPANIES = [
-    "Apple Inc.",
-    "Microsoft Corporation",
-    "Amazon.com, Inc.",
-    "Alphabet Inc.",
-    "Meta Platforms, Inc.",
-    "Tesla, Inc.",
-    "Berkshire Hathaway Inc.",
-    "NVIDIA Corporation",
-    "JPMorgan Chase & Co.",
-    "Johnson & Johnson",
-    "Walmart Inc.",
-    "Visa Inc.",
-    "Mastercard Incorporated",
-    "Procter & Gamble Co.",
-    "UnitedHealth Group Incorporated",
-    "Bank of America Corporation",
-    "Home Depot, Inc.",
-    "Nestlé S.A.",
-    "Samsung Electronics Co., Ltd.",
-    "LVMH Moët Hennessy – Louis Vuitton",
-    "ASML Holding N.V.",
-    "Exxon Mobil Corporation",
-    "Intel Corporation",
-    "Pfizer Inc.",
-    "The Coca-Cola Company",
-    "PepsiCo, Inc.",
-    "Chevron Corporation",
-    "Merck & Co., Inc.",
-    "Novartis International AG",
-    "Toyota Motor Corporation",
-    "Oracle Corporation",
-    "Cisco Systems, Inc.",
-    "Adobe Inc.",
-    "Salesforce, Inc.",
-    "Netflix, Inc.",
-    "International Business Machines Corporation (IBM)",
-    "The Walt Disney Company",
-    "HSBC Holdings plc",
-    "McDonald's Corporation",
-    "Nike, Inc.",
-    "Qualcomm Incorporated",
-    "Roche Holding AG",
-    "SAP SE",
-    "Abbott Laboratories",
-    "Costco Wholesale Corporation",
-    "Broadcom Inc.",
-    "Accenture plc",
-    "Chevron Corporation",
-    "Texas Instruments Incorporated",
-    "Unilever PLC"
-]
-
-THINGS_TO_KNOW = [
-    "Company name and branding",
-    "Founding date and history",
-    "Founders and key executives",
-    "Headquarters and global offices",
-    "Mission, vision, and values",
-    "Core products and services",
-    "Business model and revenue streams",
-    "Annual revenue and growth rate",
-    "Profit margins (gross, operating, net)",
-    "Market capitalization and valuation",
-    "Key financial ratios (P/E, ROE, ROI)",
-    "Stock price history and recent performance",
-    "Major investors and shareholder structure",
-    "Recent mergers, acquisitions, or divestitures",
-    "R&D spending and innovation pipeline",
-    "Competitive landscape and main rivals",
-    "Market share by region or segment",
-    "Customer segments and target markets",
-    "Pricing strategy and positioning",
-    "Supply chain structure and partners",
-    "Distribution channels and logistics",
-    "Marketing and advertising strategies",
-    "Brand perception and reputation",
-    "ESG (Environmental, Social, Governance) scores",
-    "Sustainability initiatives and impact",
-    "Corporate culture and employee count",
-    "Employee satisfaction and turnover rates",
-    "Leadership and governance practices",
-    "Regulatory environment and compliance record",
-    "Key risks and litigation history",
-    "Recent news, press releases, and media coverage",
-    "Patents, trademarks, and IP portfolio",
-    "Digital transformation and tech stack",
-    "Website traffic and social media metrics",
-    "Mobile app usage and customer reviews",
-    "Partnerships and strategic alliances",
-    "Future outlook and analyst recommendations",
-]
-
-JOB_TITLES = [
-    "Chief Executive Officer (CEO)",
-    "Chief Financial Officer (CFO)",
-    "Chief Operating Officer (COO)",
-    "Chief Technology Officer (CTO)",
-    "Chief Marketing Officer (CMO)",
-    "Head of Investor Relations",
-    "Director of Corporate Strategy",
-    "Business Development Manager",
-    "Product Manager",
-    "Marketing Manager",
-    "Brand Manager",
-    "Financial Analyst",
-    "Equity Research Analyst",
-    "Market Research Analyst",
-    "Consultant",
-    "Venture Capital Associate",
-    "Private Equity Associate",
-    "Operations Manager",
-    "Supply Chain Manager",
-    "Human Resources Manager",
-    "Sustainability Officer",
-    "Compliance Officer",
-    "Legal Counsel",
-    "Risk Manager",
-    "Data Analyst",
-    "IT Manager",
-    "Sales Director",
-    "Account Manager",
-    "Customer Success Manager",
-    "Public Relations Manager"
-]
-
-def _get_question():
-    return (f"I am a {random.choice(JOB_TITLES)} and I want to know {random.choice(THINGS_TO_KNOW)}"
-            f"about {random.choice(COMPANIES)}")