code-review-ai-cli 2.2.2__tar.gz → 2.2.4__tar.gz

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-review-ai-cli
-Version: 2.2.2
+Version: 2.2.4
 Summary: Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests
 License: MIT
 Keywords: code-review,ai,azure-devops,tfs,pull-request,llm
@@ -17,6 +17,8 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: requests>=2.28.0
 Requires-Dist: PyYAML>=6.0
+Requires-Dist: litellm>=1.40.0
+Requires-Dist: tiktoken>=0.7.0
 Provides-Extra: bedrock
 Requires-Dist: boto3>=1.34.0; extra == "bedrock"
 Provides-Extra: openai
@@ -29,6 +31,8 @@ Provides-Extra: tree-sitter
 Requires-Dist: tree-sitter-languages>=1.10.2; extra == "tree-sitter"
 Provides-Extra: all
 Requires-Dist: boto3>=1.34.0; extra == "all"
+Requires-Dist: litellm>=1.40.0; extra == "all"
+Requires-Dist: tiktoken>=0.7.0; extra == "all"
 Requires-Dist: openai>=1.0.0; extra == "all"
 Requires-Dist: google-generativeai>=0.3.0; extra == "all"
 Requires-Dist: anthropic>=0.18.0; extra == "all"
@@ -51,6 +55,8 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 - Configuration exclusively via `config.yaml`
 - Providers LLM: OpenAI, Azure OpenAI, Gemini, Claude, Ollama, GitHub Copilot, AWS Bedrock
 - Diff filtering by **excluded path prefixes** (`excluded_paths`) and/or **file extensions** (`file_extensions_filter`)
+- Prompt token consumption estimation with LiteLLM (`Total Tokens`)
+- Validation to ensure comments are not duplicated in re-runs (CI/CD Pipelines)
 
 ## Documentation
 
@@ -63,6 +69,7 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 
 ![Review Output Format Example](/imgs/review_output.png)
 
+
 ## Installation
 
 Install from PyPI:
@@ -144,3 +151,6 @@ output:
 - `AI Review: PR (Dry-Run)`
 - `AI Review: List Active PRs`
 - `AI Review: Interactive Mode`
+
+## License
+This project is licensed under the Apache Licence 2.0 — see the [LICENSE](/License).

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/README.md

@@ -11,6 +11,8 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 - Configuration exclusively via `config.yaml`
 - Providers LLM: OpenAI, Azure OpenAI, Gemini, Claude, Ollama, GitHub Copilot, AWS Bedrock
 - Diff filtering by **excluded path prefixes** (`excluded_paths`) and/or **file extensions** (`file_extensions_filter`)
+- Prompt token consumption estimation with LiteLLM (`Total Tokens`)
+- Validation to ensure comments are not duplicated in re-runs (CI/CD Pipelines)
 
 ## Documentation
 
@@ -23,6 +25,7 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 
 ![Review Output Format Example](/imgs/review_output.png)
 
+
 ## Installation
 
 Install from PyPI:
@@ -104,3 +107,6 @@ output:
 - `AI Review: PR (Dry-Run)`
 - `AI Review: List Active PRs`
 - `AI Review: Interactive Mode`
+
+## License
+This project is licensed under the Apache Licence 2.0 — see the [LICENSE](/License).

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/code_review_ai_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-review-ai-cli
-Version: 2.2.2
+Version: 2.2.4
 Summary: Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests
 License: MIT
 Keywords: code-review,ai,azure-devops,tfs,pull-request,llm
@@ -17,6 +17,8 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: requests>=2.28.0
 Requires-Dist: PyYAML>=6.0
+Requires-Dist: litellm>=1.40.0
+Requires-Dist: tiktoken>=0.7.0
 Provides-Extra: bedrock
 Requires-Dist: boto3>=1.34.0; extra == "bedrock"
 Provides-Extra: openai
@@ -29,6 +31,8 @@ Provides-Extra: tree-sitter
 Requires-Dist: tree-sitter-languages>=1.10.2; extra == "tree-sitter"
 Provides-Extra: all
 Requires-Dist: boto3>=1.34.0; extra == "all"
+Requires-Dist: litellm>=1.40.0; extra == "all"
+Requires-Dist: tiktoken>=0.7.0; extra == "all"
 Requires-Dist: openai>=1.0.0; extra == "all"
 Requires-Dist: google-generativeai>=0.3.0; extra == "all"
 Requires-Dist: anthropic>=0.18.0; extra == "all"
@@ -51,6 +55,8 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 - Configuration exclusively via `config.yaml`
 - Providers LLM: OpenAI, Azure OpenAI, Gemini, Claude, Ollama, GitHub Copilot, AWS Bedrock
 - Diff filtering by **excluded path prefixes** (`excluded_paths`) and/or **file extensions** (`file_extensions_filter`)
+- Prompt token consumption estimation with LiteLLM (`Total Tokens`)
+- Validation to ensure comments are not duplicated in re-runs (CI/CD Pipelines)
 
 ## Documentation
 
@@ -63,6 +69,7 @@ AI Code Review is an automated AI-powered code review CLI, compatible with vario
 
 ![Review Output Format Example](/imgs/review_output.png)
 
+
 ## Installation
 
 Install from PyPI:
@@ -144,3 +151,6 @@ output:
 - `AI Review: PR (Dry-Run)`
 - `AI Review: List Active PRs`
 - `AI Review: Interactive Mode`
+
+## License
+This project is licensed under the Apache Licence 2.0 — see the [LICENSE](/License).

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/code_review_ai_cli.egg-info/SOURCES.txt

@@ -15,6 +15,7 @@ src/git_utils.py
 src/llm_client.py
 src/prompt_utils.py
 src/tfs_client.py
+src/token_estimator.py
 src/prompts/config.yaml.template
 src/prompts/review_prompt.md.template
 tests/test_ai_review.py
@@ -24,4 +25,5 @@ tests/test_formatter.py
 tests/test_git_utils.py
 tests/test_llm_client.py
 tests/test_prompt_utils.py
-tests/test_tfs_client.py
+tests/test_tfs_client.py
+tests/test_token_estimator.py

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/code_review_ai_cli.egg-info/requires.txt

@@ -1,8 +1,12 @@
 requests>=2.28.0
 PyYAML>=6.0
+litellm>=1.40.0
+tiktoken>=0.7.0
 
 [all]
 boto3>=1.34.0
+litellm>=1.40.0
+tiktoken>=0.7.0
 openai>=1.0.0
 google-generativeai>=0.3.0
 anthropic>=0.18.0

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "code-review-ai-cli"
-version = "2.2.2"
+version = "2.2.4"
 description = "Automated AI-powered code review CLI for Azure DevOps / TFS Pull Requests"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -24,6 +24,8 @@ classifiers = [
 dependencies = [
     "requests>=2.28.0",
     "PyYAML>=6.0",
+    "litellm>=1.40.0",
+    "tiktoken>=0.7.0",
 ]
 
 [project.optional-dependencies]
@@ -34,6 +36,8 @@ claude = ["anthropic>=0.18.0"]
 tree-sitter = ["tree-sitter-languages>=1.10.2"]
 all = [
     "boto3>=1.34.0",
+    "litellm>=1.40.0",
+    "tiktoken>=0.7.0",
     "openai>=1.0.0",
     "google-generativeai>=0.3.0",
     "anthropic>=0.18.0",

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/__init__.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 # AI Code Review - Internal modules
 from importlib.metadata import version, PackageNotFoundError
 

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/ai_review.py

@@ -1,4 +1,7 @@
 #!/usr/bin/env python3
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 AI Code Review - Main Script
 ==============================
@@ -516,6 +519,16 @@ def run_pr_review_workflow(args: argparse.Namespace, config: ReviewConfig,
     elapsed = time.time() - start_time
     progress.stop(formatter.format_success(f"Review completed in {elapsed:.1f}s"))
 
+    token_estimate = getattr(llm, "last_token_estimate", None)
+    if token_estimate is not None and token_estimate.total_prompt_tokens is not None:
+        print(formatter.format_info(
+            f"Total Prompt: {token_estimate.total_prompt_tokens} tokens"
+        ))
+    else:
+        print(formatter.format_warning(
+            "Total Prompt: unavailable (LiteLLM and local tiktoken could not count the model)"
+        ))
+
     # --- Show structured comments preview ---
     print(formatter.format_structured_comments(
         structured_comments,
@@ -582,8 +595,15 @@ def run_pr_review_workflow(args: argparse.Namespace, config: ReviewConfig,
             f"---\n"
             f"*Automatic review generated by AI Code Review v{VERSION}*"
         )
-        tfs.post_general_comment(repo_name, pr_id, summary_comment)
-        print(formatter.format_success("General summary posted to PR."))
+        result = tfs.post_general_comment(
+            repo_name, pr_id, summary_comment, dedup_marker="summary"
+        )
+        if result.get("skipped"):
+            print(formatter.format_info(
+                "General summary already present, skipped."
+            ))
+        else:
+            print(formatter.format_success("General summary posted to PR."))
     except TFSError as exc:
         print(formatter.format_warning(f"Could not post general summary: {exc}"))
 

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/config.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 Configuration Module - AI Code Review
 =======================================

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/context_extractor.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 Context Extractor Module - AI Code Review
 ==========================================

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/formatter.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 Formatting Module - AI Code Review
 ====================================
@@ -294,7 +297,8 @@ class ReviewFormatter:
         lines = [f"\n{c.BOLD}📤 Comment posting results:{c.RESET}\n"]
 
         success_count = sum(1 for r in results if r.get("success"))
-        fail_count = len(results) - success_count
+        skipped_count = sum(1 for r in results if r.get("skipped"))
+        fail_count = len(results) - success_count - skipped_count
 
         for r in results:
             if r.get("success"):
@@ -319,6 +323,7 @@ class ReviewFormatter:
         lines.append(
             f"\n  {c.BOLD}Total: "
             f"{c.GREEN}{success_count} posted{c.RESET}, "
+            f"{c.YELLOW}{skipped_count} skipped{c.RESET}, "
             f"{c.RED}{fail_count} failed{c.RESET}"
         )
         return "\n".join(lines)

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/git_utils.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 Git Utilities Module - AI Code Review
 ========================================

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/llm_client.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 LLM Client Module - AI Code Review
 =====================================
@@ -18,6 +21,7 @@ import os
 
 from .config import ReviewConfig
 from .prompt_utils import detect_langs, filter_prompt_by_langs
+from .token_estimator import TokenEstimate, TokenEstimator
 
 
 class LLMError(Exception):
@@ -317,6 +321,11 @@ class LLMClient:
 
     def __init__(self, config: ReviewConfig):
         self.config = config
+        self.last_token_estimate = TokenEstimate(
+            total_prompt_tokens=None,
+            model=config.get_effective_model(),
+            error="No review has been performed yet.",
+        )
 
     def _dump_prompt_debug(self, system_prompt: str, user_message: str) -> None:
         """
@@ -420,6 +429,15 @@ class LLMClient:
         user_message = build_user_message(diff, files_summary, merged_context)
         self._dump_prompt_debug(system_prompt, user_message)
 
+        self.last_token_estimate = TokenEstimator().estimate(
+            provider=self.config.llm_provider,
+            model=self.config.get_effective_model(),
+            messages=[
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_message},
+            ],
+        )
+
         provider = self.config.llm_provider.lower()
 
         if provider == "openai":

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/prompt_utils.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 Prompt Utilities Module - AI Code Review
 ========================================

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/src/tfs_client.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """
 TFS/Azure DevOps Module - AI Code Review
 ==========================================
@@ -15,7 +18,9 @@ Works with both on-premises TFS and Azure DevOps Services.
 
 import base64
 import difflib
+import hashlib
 import os
+import re
 from typing import Optional
 
 from .config import ReviewConfig
@@ -32,6 +37,12 @@ class TFSClient:
 
     API_VERSION = "7.0"
 
+    # Hidden fingerprint marker embedded in AI-generated comments to detect
+    # and skip duplicates on pipeline re-runs. Rendered as an HTML comment,
+    # which Azure DevOps does not display in the PR UI.
+    MARKER_PREFIX = "ai-review:"
+    _MARKER_RE = re.compile(r"<!--\s*ai-review:([0-9a-zA-Z_]+)\s*-->")
+
     def __init__(self, config: ReviewConfig):
         self.config = config
         self.base_url = config.tfs_base_url.rstrip("/")
@@ -614,20 +625,36 @@ class TFSClient:
     # Pull Requests - Comments
     # ==================================================================
     def post_general_comment(self, repository: str, pr_id: int,
-                             comment: str, status: str = "active") -> dict:
+                             comment: str, status: str = "active",
+                             dedup_marker: Optional[str] = None) -> dict:
         """
         Posts a general comment on a Pull Request (not associated with a file).
-        
+
         Args:
             repository: Repository name.
             pr_id: Pull Request ID.
             comment: Comment text (supports Markdown).
             status: Thread status - "active", "fixed", "wontFix",
                     "closed", "pending", "byDesign"
-            
+            dedup_marker: Optional stable identifier. When provided, a hidden
+                          fingerprint is appended to the comment and existing PR
+                          threads are checked first; if the same marker already
+                          exists, the comment is skipped instead of duplicated.
+
         Returns:
-            Created thread data.
+            Created thread data, or a ``{"skipped": True, ...}`` dict when the
+            comment was skipped due to an already existing marker.
         """
+        marker: Optional[str] = None
+        if dedup_marker is not None:
+            marker = self._general_marker(dedup_marker)
+            existing = self._extract_existing_markers(
+                self.list_pr_threads(repository, pr_id)
+            )
+            if marker in existing:
+                return {"skipped": True, "marker": marker}
+            comment = f"{comment}\n\n{self._marker_html(marker)}"
+
         path = f"git/repositories/{repository}/pullrequests/{pr_id}/threads"
         data = {
             "comments": [
@@ -726,13 +753,29 @@ class TFSClient:
         comment_mode = (comment_mode or "structured").lower()
         use_inline_comments = comment_mode == "structured"
 
-        for c in comments:
-            # Build formatted comment text
-            text = self._format_review_comment(c)
+        # Load existing markers once to skip comments already posted in a
+        # previous run (deduplication is always active).
+        existing_markers = self._extract_existing_markers(
+            self.list_pr_threads(repository, pr_id)
+        )
 
+        for c in comments:
             file_path = c.get("file", "")
             line = c.get("line", 0)
 
+            marker = self._comment_marker(c)
+            if marker in existing_markers:
+                results.append({
+                    "skipped": True,
+                    "file": file_path,
+                    "line": line,
+                    "error": "duplicate (already posted)",
+                })
+                continue
+
+            # Build formatted comment text with the hidden fingerprint appended.
+            text = f"{self._format_review_comment(c)}\n\n{self._marker_html(marker)}"
+
             try:
                 if use_inline_comments and file_path and line > 0:
                     # Inline comment
@@ -744,6 +787,8 @@ class TFSClient:
                     result = self.post_general_comment(
                         repository, pr_id, text
                     )
+                # Record the marker to avoid duplicates within the same batch.
+                existing_markers.add(marker)
                 results.append({
                     "success": True,
                     "file": file_path,
@@ -790,6 +835,86 @@ class TFSClient:
 
         return "\n".join(parts)
 
+    # ==================================================================
+    # Deduplication - hidden fingerprint markers
+    # ==================================================================
+    def list_pr_threads(self, repository: str, pr_id: int) -> list[dict]:
+        """
+        Lists all comment threads of a Pull Request.
+
+        Args:
+            repository: Repository name.
+            pr_id: Pull Request ID.
+
+        Returns:
+            List of thread dictionaries (empty list when none exist).
+        """
+        path = f"git/repositories/{repository}/pullrequests/{pr_id}/threads"
+        response = self._get(path)
+        return response.get("value", [])
+
+    def _comment_marker(self, comment: dict) -> str:
+        """
+        Builds a stable fingerprint for a structured review comment.
+
+        The hash is derived from the file, line and type so the same finding
+        location produces the same marker across runs, regardless of small
+        wording changes in the comment text.
+
+        Args:
+            comment: Structured comment with ``file``, ``line`` and ``type`` keys.
+
+        Returns:
+            Hexadecimal SHA-256 digest string.
+        """
+        payload = "|".join([
+            str(comment.get("file", "")),
+            str(comment.get("line", "")),
+            str(comment.get("type", "")),
+        ])
+        return hashlib.sha256(payload.encode("utf-8")).hexdigest()
+
+    def _general_marker(self, identifier: str) -> str:
+        """
+        Builds a stable marker for a general comment from a fixed identifier.
+
+        Args:
+            identifier: Stable identifier (e.g. ``"summary"``).
+
+        Returns:
+            Sanitized marker string usable inside the HTML comment.
+        """
+        return re.sub(r"[^0-9a-zA-Z_]", "_", identifier)
+
+    def _marker_html(self, marker: str) -> str:
+        """
+        Wraps a marker into a hidden HTML comment.
+
+        Args:
+            marker: Fingerprint value.
+
+        Returns:
+            HTML comment string invisible in the rendered PR.
+        """
+        return f"<!-- {self.MARKER_PREFIX}{marker} -->"
+
+    def _extract_existing_markers(self, threads: list[dict]) -> set[str]:
+        """
+        Extracts all fingerprint markers present in existing PR threads.
+
+        Args:
+            threads: Threads as returned by :meth:`list_pr_threads`.
+
+        Returns:
+            Set of marker values found across every comment.
+        """
+        markers: set[str] = set()
+        for thread in threads:
+            for comment in thread.get("comments", []) or []:
+                content = comment.get("content", "") or ""
+                markers.update(self._MARKER_RE.findall(content))
+        return markers
+
     def _status_to_int(self, status: str) -> int:
         """Converts status string to API integer."""
         status_map = {

code_review_ai_cli-2.2.4/src/token_estimator.py

@@ -0,0 +1,98 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
+from __future__ import annotations
+from dataclasses import dataclass
+
+
+@dataclass(frozen=True)
+class TokenEstimate:
+    """Result of estimating the input tokens for an LLM request."""
+
+    total_prompt_tokens: int | None
+    model: str
+    error: str = ""
+
+
+class TokenEstimator:
+    """Estimates prompt tokens with LiteLLM for configured providers."""
+
+    _PROVIDER_PREFIXES = {
+        "azure_openai": "azure",
+        "gemini": "gemini",
+        "claude": "anthropic",
+        "ollama": "ollama",
+        "copilot": "github",
+        "bedrock": "bedrock",
+    }
+
+    def estimate(
+        self,
+        provider: str,
+        model: str,
+        messages: list[dict[str, str]],
+    ) -> TokenEstimate:
+        """Estimates input tokens, falling back to the unqualified model.
+
+        Args:
+            provider: Provider configured for the LLM request.
+            model: Effective model name, including configured defaults.
+            messages: Complete chat payload sent to the provider.
+
+        Returns:
+            A token estimate, or an unavailable result when LiteLLM cannot
+            resolve either the provider-qualified or raw model name.
+        """
+        qualified_model = self._qualified_model(provider, model)
+        errors: list[str] = []
+
+        for candidate in dict.fromkeys((qualified_model, model)):
+            try:
+                token_count = self._count(candidate, messages)
+                return TokenEstimate(token_count, candidate)
+            except Exception as exc:
+                errors.append(f"{candidate}: {exc}")
+
+        try:
+            token_count = self._count_locally(model, messages)
+            return TokenEstimate(token_count, "local/tiktoken")
+        except Exception as exc:
+            errors.append(f"local/tiktoken: {exc}")
+
+        return TokenEstimate(
+            total_prompt_tokens=None,
+            model=qualified_model,
+            error="; ".join(errors),
+        )
+
+    @classmethod
+    def _qualified_model(cls, provider: str, model: str) -> str:
+        """Builds the LiteLLM model name for a configured provider."""
+        normalized_provider = provider.lower().strip()
+        prefix = cls._PROVIDER_PREFIXES.get(normalized_provider)
+        if not prefix:
+            return model
+        return f"{prefix}/{model}"
+
+    @staticmethod
+    def _count(model: str, messages: list[dict[str, str]]) -> int:
+        """Calls LiteLLM lazily so importing the CLI remains lightweight."""
+        from litellm import token_counter
+
+        return int(token_counter(model=model, messages=messages))
+
+    @staticmethod
+    def _count_locally(model: str, messages: list[dict[str, str]]) -> int:
+        """Counts message text locally when LiteLLM has no model support."""
+        import tiktoken
+
+        try:
+            encoding = tiktoken.encoding_for_model(model)
+        except KeyError:
+            encoding = tiktoken.get_encoding("cl100k_base")
+
+        serialized_messages = "\n".join(
+            f"{message.get('role', '')}: {message.get('content', '')}"
+            for message in messages
+        )
+        return len(encoding.encode(serialized_messages))

{code_review_ai_cli-2.2.2 → code_review_ai_cli-2.2.4}/tests/test_ai_review.py

@@ -1,3 +1,6 @@
+# License: Apache-2.0
+# Copyright 2026 Sérgio Sereno
+
 """Tests for the AI review CLI workflow."""
 
 from __future__ import annotations