contextpress 0.5.0__tar.gz → 0.5.2__tar.gz

{contextpress-0.5.0 → contextpress-0.5.2}/CHANGELOG.md

@@ -4,6 +4,16 @@ All notable changes to `contextpress` are recorded here.
 The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/)
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.5.2] - 2026-07-22
+
+- **`OpenAIBackend`** — optional `api_key` / env (`OPENAI_API_KEY`); no manual client required.
+
+## [0.5.1] - 2026-07-22
+
+- **`ClaudeBackend`** — shortcut for Anthropic Claude (`ANTHROPIC_API_KEY` or `api_key=`).
+- **`GeminiBackend`** — optional `model_name` + `api_key` / env (`GOOGLE_API_KEY`, `GEMINI_API_KEY`).
+- Examples: `examples/llm_tier_claude.py`, `examples/llm_tier_gemini.py`.
+
 ## [0.5.0] - 2026-07-22
 
 - **`compress_async()`** — async wrapper around ``compress()`` for asyncio apps.

{contextpress-0.5.0 → contextpress-0.5.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: contextpress
-Version: 0.5.0
+Version: 0.5.2
 Summary: Deterministic context compression for LLM chat, RAG, and agent pipelines
 Project-URL: Homepage, https://github.com/Taha-azizi/contextpress
 Project-URL: Documentation, https://github.com/Taha-azizi/contextpress#readme
@@ -401,7 +401,7 @@ ContextManager(type="agent")
 | **Techniques** | Rules, TF–IDF, cosine similarity, NLTK, Sumy extractive summarization, tiktoken | Your provider’s chat/completions API (you supply the client) |
 | **API key** | None | Required for your chosen provider (OpenAI, Anthropic, …) |
 | **Determinism** | Deterministic for a fixed input and settings | Non-deterministic (model sampling) |
-| **How to enable** | Default: `ContextManager()` runs Tier 1 only | Pass `llm_backend=` (`OpenAIBackend`, `AnthropicBackend`, `GeminiBackend`, **`OllamaBackend`**, or custom `LLMBackend`) |
+| **How to enable** | Default: `ContextManager()` runs Tier 1 only | Pass `llm_backend=` (`OpenAIBackend`, **`ClaudeBackend`**, `GeminiBackend`, `OllamaBackend`, or custom `LLMBackend`) |
 
 **Note:** `ContextManager(model="gpt-4")` is only for **tiktoken** encoding when counting tokens in the **budget** stage. It does **not** call that model unless you also pass **`llm_backend`**.
 
@@ -459,11 +459,10 @@ Optional constructor knobs: **`llm_min_input_chars`**, **`llm_max_summary_tokens
 **Install SDKs** (not bundled): `pip install openai`, `anthropic`, and/or **`ollama`** (for local Ollama), or `pip install "contextpress[llm]"` from this repo’s `pyproject.toml` to pull all optional LLM clients.
 
 ```python
-from openai import OpenAI
 from contextpress import ContextManager
 from contextpress.llm.adapters import OpenAIBackend
 
-backend = OpenAIBackend(client=OpenAI(), model="gpt-4o-mini")
+backend = OpenAIBackend(model="gpt-4o-mini")  # uses OPENAI_API_KEY
 cm = ContextManager(
     type="chat",
     llm_backend=backend,
@@ -494,6 +493,32 @@ out = cm.compress(messages, token_budget=4000)
 
 Runnable script: [`examples/llm_tier_ollama.py`](examples/llm_tier_ollama.py).
 
+**Claude (Anthropic)** — `pip install anthropic`, set `ANTHROPIC_API_KEY`:
+
+```python
+from contextpress import ContextManager
+from contextpress.llm.adapters import ClaudeBackend
+
+backend = ClaudeBackend(model="claude-haiku-4-5")
+cm = ContextManager(type="chat", llm_backend=backend, llm_min_input_chars=500)
+out = cm.compress(messages, token_budget=4000)
+```
+
+Runnable script: [`examples/llm_tier_claude.py`](examples/llm_tier_claude.py).
+
+**Gemini (Google)** — `pip install google-generativeai`, set `GOOGLE_API_KEY`:
+
+```python
+from contextpress import ContextManager
+from contextpress.llm.adapters import GeminiBackend
+
+backend = GeminiBackend(model_name="gemini-2.0-flash")
+cm = ContextManager(type="chat", llm_backend=backend, llm_min_input_chars=500)
+out = cm.compress(messages, token_budget=4000)
+```
+
+Runnable script: [`examples/llm_tier_gemini.py`](examples/llm_tier_gemini.py).
+
 ```bash
 pip install ollama
 ollama pull llama3.2

{contextpress-0.5.0 → contextpress-0.5.2}/README.md

@@ -160,7 +160,7 @@ ContextManager(type="agent")
 | **Techniques** | Rules, TF–IDF, cosine similarity, NLTK, Sumy extractive summarization, tiktoken | Your provider’s chat/completions API (you supply the client) |
 | **API key** | None | Required for your chosen provider (OpenAI, Anthropic, …) |
 | **Determinism** | Deterministic for a fixed input and settings | Non-deterministic (model sampling) |
-| **How to enable** | Default: `ContextManager()` runs Tier 1 only | Pass `llm_backend=` (`OpenAIBackend`, `AnthropicBackend`, `GeminiBackend`, **`OllamaBackend`**, or custom `LLMBackend`) |
+| **How to enable** | Default: `ContextManager()` runs Tier 1 only | Pass `llm_backend=` (`OpenAIBackend`, **`ClaudeBackend`**, `GeminiBackend`, `OllamaBackend`, or custom `LLMBackend`) |
 
 **Note:** `ContextManager(model="gpt-4")` is only for **tiktoken** encoding when counting tokens in the **budget** stage. It does **not** call that model unless you also pass **`llm_backend`**.
 
@@ -218,11 +218,10 @@ Optional constructor knobs: **`llm_min_input_chars`**, **`llm_max_summary_tokens
 **Install SDKs** (not bundled): `pip install openai`, `anthropic`, and/or **`ollama`** (for local Ollama), or `pip install "contextpress[llm]"` from this repo’s `pyproject.toml` to pull all optional LLM clients.
 
 ```python
-from openai import OpenAI
 from contextpress import ContextManager
 from contextpress.llm.adapters import OpenAIBackend
 
-backend = OpenAIBackend(client=OpenAI(), model="gpt-4o-mini")
+backend = OpenAIBackend(model="gpt-4o-mini")  # uses OPENAI_API_KEY
 cm = ContextManager(
     type="chat",
     llm_backend=backend,
@@ -253,6 +252,32 @@ out = cm.compress(messages, token_budget=4000)
 
 Runnable script: [`examples/llm_tier_ollama.py`](examples/llm_tier_ollama.py).
 
+**Claude (Anthropic)** — `pip install anthropic`, set `ANTHROPIC_API_KEY`:
+
+```python
+from contextpress import ContextManager
+from contextpress.llm.adapters import ClaudeBackend
+
+backend = ClaudeBackend(model="claude-haiku-4-5")
+cm = ContextManager(type="chat", llm_backend=backend, llm_min_input_chars=500)
+out = cm.compress(messages, token_budget=4000)
+```
+
+Runnable script: [`examples/llm_tier_claude.py`](examples/llm_tier_claude.py).
+
+**Gemini (Google)** — `pip install google-generativeai`, set `GOOGLE_API_KEY`:
+
+```python
+from contextpress import ContextManager
+from contextpress.llm.adapters import GeminiBackend
+
+backend = GeminiBackend(model_name="gemini-2.0-flash")
+cm = ContextManager(type="chat", llm_backend=backend, llm_min_input_chars=500)
+out = cm.compress(messages, token_budget=4000)
+```
+
+Runnable script: [`examples/llm_tier_gemini.py`](examples/llm_tier_gemini.py).
+
 ```bash
 pip install ollama
 ollama pull llama3.2

{contextpress-0.5.0 → contextpress-0.5.2}/contextpress/__init__.py

@@ -16,7 +16,7 @@ __all__ = [
     "CompressionResult",
     "CompressionStats",
 ]
-__version__ = "0.5.0"
+__version__ = "0.5.2"
 
 
 def __getattr__(name: str) -> Any:

{contextpress-0.5.0 → contextpress-0.5.2}/contextpress/llm/__init__.py

@@ -1,5 +1,6 @@
 from contextpress.llm.adapters import (
     AnthropicBackend,
+    ClaudeBackend,
     GeminiBackend,
     OllamaBackend,
     OpenAIBackend,
@@ -12,6 +13,7 @@ __all__ = [
     "OpenAIBackend",
     "OpenAICompatibleBackend",
     "AnthropicBackend",
+    "ClaudeBackend",
     "GeminiBackend",
     "OllamaBackend",
 ]

{contextpress-0.5.0 → contextpress-0.5.2}/contextpress/llm/adapters.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import os
 import warnings
 from typing import Any
 
@@ -26,20 +27,45 @@ def _ollama_response_text(resp: Any) -> str:
 
 class OpenAIBackend(LLMBackend):
     """
-    Adapter for OpenAI-compatible APIs.
-    Requires: pip install openai
-    User must pass their own client instance.
+    Adapter for **OpenAI** chat APIs.
+
+    Requires: ``pip install openai``
+
+    Usage (API key from env ``OPENAI_API_KEY`` or ``api_key=``)::
 
-    Usage:
-        from openai import OpenAI
         from contextpress.llm.adapters import OpenAIBackend
 
-        backend = OpenAIBackend(client=OpenAI(), model="gpt-4o-mini")
-        cm = ContextManager(type="chat", llm_backend=backend)
+        backend = OpenAIBackend(model=\"gpt-4o-mini\")
+        cm = ContextManager(type=\"chat\", llm_backend=backend)
+
+    Or pass your own client::
+
+        from openai import OpenAI
+
+        backend = OpenAIBackend(client=OpenAI(), model=\"gpt-4o-mini\")
     """
 
-    def __init__(self, client: Any, model: str = "gpt-4o-mini"):
-        self.client = client
+    def __init__(
+        self,
+        client: Any | None = None,
+        model: str = "gpt-4o-mini",
+        *,
+        api_key: str | None = None,
+    ):
+        if client is not None:
+            self.client = client
+            self.model = model
+            return
+        try:
+            from openai import OpenAI
+        except ImportError as exc:
+            raise ImportError(
+                "OpenAIBackend requires the 'openai' package. " "Install with: pip install openai"
+            ) from exc
+        key = api_key or os.environ.get("OPENAI_API_KEY")
+        if not key:
+            raise ValueError("Set OPENAI_API_KEY or pass api_key= for OpenAIBackend")
+        self.client = OpenAI(api_key=key)
         self.model = model
 
     def summarize(self, text: str, max_tokens: int) -> str:
@@ -135,6 +161,43 @@ class AnthropicBackend(LLMBackend):
             return list(range(len(turns)))
 
 
+class ClaudeBackend(AnthropicBackend):
+    """
+    Convenience adapter for **Anthropic Claude** (wraps ``AnthropicBackend``).
+
+    Requires: ``pip install anthropic``
+
+    Usage::
+
+        from contextpress.llm.adapters import ClaudeBackend
+
+        backend = ClaudeBackend(model=\"claude-haiku-4-5\")  # uses ANTHROPIC_API_KEY
+        cm = ContextManager(type=\"chat\", llm_backend=backend)
+    """
+
+    def __init__(
+        self,
+        model: str = "claude-haiku-4-5",
+        *,
+        api_key: str | None = None,
+        client: Any | None = None,
+    ):
+        if client is not None:
+            super().__init__(client=client, model=model)
+            return
+        try:
+            import anthropic
+        except ImportError as exc:
+            raise ImportError(
+                "ClaudeBackend requires the 'anthropic' package. "
+                "Install with: pip install anthropic"
+            ) from exc
+        key = api_key or os.environ.get("ANTHROPIC_API_KEY")
+        if not key:
+            raise ValueError("Set ANTHROPIC_API_KEY or pass api_key= for ClaudeBackend")
+        super().__init__(client=anthropic.Anthropic(api_key=key), model=model)
+
+
 class OllamaBackend(LLMBackend):
     """
     Adapter for **Ollama** (local or remote) using the official ``ollama`` Python library.
@@ -270,18 +333,41 @@ class GeminiBackend(LLMBackend):
 
     Requires: ``pip install google-generativeai``
 
-    Usage::
+    Usage (API key from env ``GOOGLE_API_KEY`` or ``GEMINI_API_KEY``)::
 
-        import google.generativeai as genai
         from contextpress.llm.adapters import GeminiBackend
 
+        backend = GeminiBackend(model_name=\"gemini-2.0-flash\")
+        cm = ContextManager(type=\"chat\", llm_backend=backend)
+
+    Or pass a preconfigured model instance::
+
+        import google.generativeai as genai
         genai.configure(api_key=\"...\")
         backend = GeminiBackend(model=genai.GenerativeModel(\"gemini-2.0-flash\"))
-        cm = ContextManager(type=\"chat\", llm_backend=backend)
     """
 
-    def __init__(self, model: Any):
-        self.model = model
+    def __init__(
+        self,
+        model: Any | None = None,
+        *,
+        model_name: str = "gemini-2.0-flash",
+        api_key: str | None = None,
+    ):
+        if model is not None:
+            self.model = model
+            return
+        try:
+            import google.generativeai as genai
+        except ImportError as exc:
+            raise ImportError(
+                "GeminiBackend requires 'google-generativeai'. "
+                "Install with: pip install google-generativeai"
+            ) from exc
+        key = api_key or os.environ.get("GOOGLE_API_KEY") or os.environ.get("GEMINI_API_KEY")
+        if key:
+            genai.configure(api_key=key)
+        self.model = genai.GenerativeModel(model_name)
 
     def _generate(self, prompt: str, *, max_tokens: int) -> str:
         resp = self.model.generate_content(

contextpress-0.5.2/examples/llm_tier_claude.py

@@ -0,0 +1,50 @@
+"""
+Tier 2 compression with Anthropic Claude.
+
+Requires ``pip install anthropic`` and ``ANTHROPIC_API_KEY``.
+
+Run:
+    set ANTHROPIC_API_KEY=sk-ant-...   # Windows
+    python examples/llm_tier_claude.py
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+
+from contextpress import ContextManager
+from contextpress.llm.adapters import ClaudeBackend
+
+
+def main() -> None:
+    if not os.environ.get("ANTHROPIC_API_KEY"):
+        print("Set ANTHROPIC_API_KEY to run this example.", file=sys.stderr)
+        sys.exit(1)
+
+    messages = [
+        {"role": "system", "content": "You are a concise assistant."},
+        {"role": "user", "content": "What is Python? " + ("Explain briefly. " * 200)},
+        {"role": "assistant", "content": "Python is a programming language. " * 150},
+        {"role": "user", "content": "And numpy? " * 80},
+    ]
+
+    backend = ClaudeBackend(model="claude-haiku-4-5")
+    cm = ContextManager(
+        type="chat",
+        llm_backend=backend,
+        compression="medium",
+        llm_min_input_chars=500,
+        llm_max_summary_tokens=512,
+    )
+
+    out = cm.compress(messages, token_budget=2000)
+    print("turns in:", len(messages), "out:", len(out))
+    for i, m in enumerate(out):
+        c = m["content"]
+        preview = (c[:200] + "…") if isinstance(c, str) and len(c) > 200 else c
+        print(i, m["role"] + ":", preview)
+
+
+if __name__ == "__main__":
+    main()

contextpress-0.5.2/examples/llm_tier_gemini.py

@@ -0,0 +1,50 @@
+"""
+Tier 2 compression with Google Gemini.
+
+Requires ``pip install google-generativeai`` and ``GOOGLE_API_KEY``.
+
+Run:
+    set GOOGLE_API_KEY=...   # or GEMINI_API_KEY
+    python examples/llm_tier_gemini.py
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+
+from contextpress import ContextManager
+from contextpress.llm.adapters import GeminiBackend
+
+
+def main() -> None:
+    if not (os.environ.get("GOOGLE_API_KEY") or os.environ.get("GEMINI_API_KEY")):
+        print("Set GOOGLE_API_KEY or GEMINI_API_KEY to run this example.", file=sys.stderr)
+        sys.exit(1)
+
+    messages = [
+        {"role": "system", "content": "You are a concise assistant."},
+        {"role": "user", "content": "What is Python? " + ("Explain briefly. " * 200)},
+        {"role": "assistant", "content": "Python is a programming language. " * 150},
+        {"role": "user", "content": "And numpy? " * 80},
+    ]
+
+    backend = GeminiBackend(model_name="gemini-2.0-flash")
+    cm = ContextManager(
+        type="chat",
+        llm_backend=backend,
+        compression="medium",
+        llm_min_input_chars=500,
+        llm_max_summary_tokens=512,
+    )
+
+    out = cm.compress(messages, token_budget=2000)
+    print("turns in:", len(messages), "out:", len(out))
+    for i, m in enumerate(out):
+        c = m["content"]
+        preview = (c[:200] + "…") if isinstance(c, str) and len(c) > 200 else c
+        print(i, m["role"] + ":", preview)
+
+
+if __name__ == "__main__":
+    main()

{contextpress-0.5.0 → contextpress-0.5.2}/examples/llm_tier_openai.py

@@ -17,8 +17,6 @@ from __future__ import annotations
 import os
 import sys
 
-from openai import OpenAI
-
 from contextpress import ContextManager
 from contextpress.llm.adapters import OpenAIBackend
 
@@ -35,7 +33,7 @@ def main() -> None:
         {"role": "user", "content": "And numpy? " * 80},
     ]
 
-    backend = OpenAIBackend(client=OpenAI(), model="gpt-4o-mini")
+    backend = OpenAIBackend(model="gpt-4o-mini")
     cm = ContextManager(
         type="chat",
         llm_backend=backend,

{contextpress-0.5.0 → contextpress-0.5.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "contextpress"
-version = "0.5.0"
+version = "0.5.2"
 description = "Deterministic context compression for LLM chat, RAG, and agent pipelines"
 readme = "README.md"
 license = { file = "LICENSE" }

contextpress-0.5.2/tests/test_v051.py

@@ -0,0 +1,50 @@
+import os
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from contextpress.llm.adapters import ClaudeBackend, GeminiBackend
+
+
+def test_claude_backend_with_client():
+    client = MagicMock()
+    backend = ClaudeBackend(model="claude-haiku-4-5", client=client)
+    assert backend.client is client
+    assert backend.model == "claude-haiku-4-5"
+
+
+def test_claude_backend_requires_api_key():
+    mock_anthropic = MagicMock()
+    with patch.dict(os.environ, {}, clear=True):
+        with patch.dict("sys.modules", {"anthropic": mock_anthropic}):
+            with pytest.raises(ValueError, match="ANTHROPIC_API_KEY"):
+                ClaudeBackend()
+
+
+def test_claude_backend_from_api_key():
+    mock_module = MagicMock()
+    mock_client = MagicMock()
+    mock_module.Anthropic.return_value = mock_client
+    with patch.dict(os.environ, {"ANTHROPIC_API_KEY": "sk-test"}):
+        with patch.dict("sys.modules", {"anthropic": mock_module}):
+            backend = ClaudeBackend(model="claude-haiku-4-5")
+    assert backend.client is mock_client
+
+
+def test_gemini_backend_with_model_instance():
+    model = MagicMock()
+    backend = GeminiBackend(model=model)
+    assert backend.model is model
+
+
+def test_gemini_backend_from_model_name():
+    mock_genai = MagicMock()
+    mock_model = MagicMock()
+    mock_genai.GenerativeModel.return_value = mock_model
+    fake_google = MagicMock()
+    fake_google.generativeai = mock_genai
+    with patch.dict(os.environ, {"GOOGLE_API_KEY": "key-test"}):
+        with patch.dict("sys.modules", {"google": fake_google, "google.generativeai": mock_genai}):
+            backend = GeminiBackend(model_name="gemini-2.0-flash")
+    assert backend.model is mock_model
+    mock_genai.configure.assert_called_once_with(api_key="key-test")

contextpress-0.5.2/tests/test_v052.py

@@ -0,0 +1,32 @@
+import os
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from contextpress.llm.adapters import OpenAIBackend
+
+
+def test_openai_backend_with_client():
+    client = MagicMock()
+    backend = OpenAIBackend(client=client, model="gpt-4o-mini")
+    assert backend.client is client
+    assert backend.model == "gpt-4o-mini"
+
+
+def test_openai_backend_requires_api_key():
+    mock_openai = MagicMock()
+    with patch.dict(os.environ, {}, clear=True):
+        with patch.dict("sys.modules", {"openai": mock_openai}):
+            with pytest.raises(ValueError, match="OPENAI_API_KEY"):
+                OpenAIBackend()
+
+
+def test_openai_backend_from_api_key():
+    mock_module = MagicMock()
+    mock_client = MagicMock()
+    mock_module.OpenAI.return_value = mock_client
+    with patch.dict(os.environ, {"OPENAI_API_KEY": "sk-test"}):
+        with patch.dict("sys.modules", {"openai": mock_module}):
+            backend = OpenAIBackend(model="gpt-4o-mini")
+    assert backend.client is mock_client
+    mock_module.OpenAI.assert_called_once_with(api_key="sk-test")