feat: Add A2A Translator Agent with text and URL translation capabili… - #111
Open
Pasindu-Jayasundara wants to merge 1 commit into
Open
feat: Add A2A Translator Agent with text and URL translation capabili…#111Pasindu-Jayasundara wants to merge 1 commit into
Pasindu-Jayasundara wants to merge 1 commit into
Conversation
…ties - Implemented OpenAIAgentExecutor for handling OpenAI-based agent execution. - Created TranslatorToolset for translating text and web content. - Added models for translation requests and responses using Pydantic. - Developed methods for translating text, extracting content from URLs, and detecting languages. - Set up Dockerfile and docker-compose for easy deployment. - Included README with installation instructions, usage examples, and configuration details. - Added .gitignore to exclude unnecessary files from version control. - Created supporting files and structure for the A2A Translator Agent.
There was a problem hiding this comment.
Pull request overview
This PR introduces three new A2A-compatible agents under agents/ (Translator, GitHub, Compliance Checker), each shipped with its own src/ Python package, Dockerfile, docker-compose.yml, AgentCard.json, README and pyproject.toml. It also disables the healthcheck for an existing service in docker-compose.local.yml and renames the root virtual project package in uv.lock from arithmic-nasiko to nasiko. The PR description only mentions the Translator agent, but the diff is much broader than that.
Changes:
- Add A2A Translator agent (OpenAI-driven executor, Google-Translate-backed toolset, language detection).
- Add A2A GitHub agent and A2A Compliance Checker agent with similar OpenAI executor scaffolding.
- Disable healthcheck for an existing service in
docker-compose.local.ymland rename the root project package inuv.lock.
Reviewed changes
Copilot reviewed 37 out of 41 changed files in this pull request and generated 21 comments.
Show a summary per file
| File | Description |
|---|---|
| uv.lock | Renames root virtual project arithmic-nasiko → nasiko; unrelated to PR description |
| docker-compose.local.yml | Disables healthcheck for an existing service (regression) |
| agents/a2a-translator/v1.0.0/src/translator_toolset.py | Translator tooling using Google's internal endpoint; mixed sync/async API |
| agents/a2a-translator/v1.0.0/src/openai_agent_executor.py | OpenAI tool-calling loop; has duplicate-complete() bug and schema-inference limitations |
| agents/a2a-translator/v1.0.0/src/openai_agent.py | System prompt + tool registration |
| agents/a2a-translator/v1.0.0/src/main.py | Server entrypoint; references likely-nonexistent default model IDs |
| agents/a2a-translator/v1.0.0/{Dockerfile,docker-compose.yml,AgentCard.json,pyproject.toml,README.md,run_with_phoenix.sh,.gitignore} | Deployment + docs; placeholder Phoenix key, unused googletrans dep, wrong outputModes |
| agents/a2a-github-agent/v1.0.0/src/main.py | Entrypoint; imports non-existent poc_obser and has flawed provider selection |
| agents/a2a-github-agent/v1.0.0/src/github_toolset.py | PyGithub-based tooling; naive vs. aware datetime comparison bug |
| agents/a2a-github-agent/v1.0.0/src/openai_agent_executor.py / openai_agent.py | Same executor pattern as translator |
| agents/a2a-github-agent/v1.0.0/{Dockerfile,docker-compose.yml,AgentCard.json,pyproject.toml,README.md,run_with_phoenix.sh,.gitignore} | Deployment + docs; malformed environment: in compose, README points to wrong repo |
| agents/a2a-compliance-checker/v1.0.0/src/policy_agent.py | References missing BaseAgent / base_agent module — agent won't start |
| agents/a2a-compliance-checker/v1.0.0/src/compliance_toolset.py | Shared "a2a_session" and shared document_text across requests |
| agents/a2a-compliance-checker/v1.0.0/src/agent.py | Leftover translation agent code in compliance package |
| agents/a2a-compliance-checker/v1.0.0/src/openai_agent_executor.py / openai_agent.py / main.py / models.py / tools.py | Executor + scaffolding mirroring the other agents |
| agents/a2a-compliance-checker/v1.0.0/{Dockerfile,docker-compose.yml,AgentCard.json,pyproject.toml,README.md,.gitignore} | Deployment + docs; project name compliance-checker2, port mismatch with AgentCard |
Comments suppressed due to low confidence (5)
agents/a2a-translator/v1.0.0/src/translator_toolset.py:299
translate_textisasync, buttranslate_urlanddetect_languageare sync and perform blocking I/O (requests.get, BeautifulSoup parsing, blocking call to Google Translate). When called from within the async_process_requestloop inopenai_agent_executor.py(line 110:result = method(**function_args)), these will block the event loop for the duration of the network request, defeating the async server. Either make them async (and offload viarun_in_executor, liketranslate_textdoes), or have the executor consistently dispatch sync tools to a thread.
agents/a2a-translator/v1.0.0/src/openai_agent_executor.py:221_extract_function_schemadoes not consult typing generics orOptional/X | None, so parameters likesource_language: str | None = Noneare typed"string"in the OpenAI schema but never marked as required (correct), while a list/dict parameter declared via PEP 585 generics (list[str],dict[str, Any]) would be typed as the default"string"becauseparam.annotation == listis False forlist[str]. For the current toolsets parameters are mostlystr/int, but as new tools get added this will silently produce wrong schemas. Usetyping.get_origin/get_argsfor proper inference.
agents/a2a-translator/v1.0.0/src/openai_agent_executor.py:181inspectis already imported at the top of the file (line 3). Re-importing it inside this method shadows nothing but is unnecessary and out of place; drop the innerimport inspect.
agents/a2a-translator/v1.0.0/src/openai_agent_executor.py:177- When
max_iterationsis reached, this branch sends a final error artifact and callstask_updater.complete(), but the prior iteration of the loop already calledtask_updater.complete()(line 156) on a successful path is fine — however if the loop exits thewhilenormally viabreakafter the success path on line 157,iterationis still<= max_iterations, yet thisif iteration >= max_iterations(line 170) will beTruewhenever the loop completed its 10th iteration successfully, causing a duplicatecomplete()and a spurious "exceeded the maximum number of iterations" message after a valid final answer. Useelseon thewhileloop, or set acompletedflag.
agents/a2a-translator/v1.0.0/src/translator_toolset.py:251 - Translation result text is truncated to 5000 characters via slicing before being sent for translation (line 250-251 appends a literal
"..."to whatever happens to be the 5000th character — potentially mid-byte in a multi-byte UTF-8 sequence when the underlying string was just decoded). Pythonstrslicing is by codepoints so it won't break encoding, but it will routinely cut in the middle of a sentence/word and feed that to Google Translate, producing low-quality output and a trailing literal"..."in the source language. Consider sentence-aware truncation, or surfacing a warning to the user that the page was truncated.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+3
to
+22
| from base_agent import BaseAgent | ||
|
|
||
| logger = logging.getLogger(__name__) | ||
|
|
||
|
|
||
| class PolicyAgent(BaseAgent): | ||
| """Handles LLM-based Policy Inquiries""" | ||
|
|
||
| def get_response(self, query: str, session_id: str) -> str: | ||
| logger.info( | ||
| f"PolicyAgent generating response for session {session_id}, query='{query}'" | ||
| ) | ||
| system_prompt = f""" | ||
| You are a specialized Policy Compliance Checker Agent. | ||
| Your expertise is analyzing documents and content for policy violations and compliance issues. | ||
|
|
||
| DOCUMENT UNDER REVIEW: | ||
| \"\"\" | ||
| {self.document_parser.document_text} | ||
| \"\"\" |
Comment on lines
+21
to
+24
| from poc_obser.tracing_utils import bootstrap_tracing | ||
|
|
||
| bootstrap_tracing(project_name="a2a-github-agent") | ||
|
|
Comment on lines
+1
to
+76
| """ | ||
| Core agent logic for translation. | ||
| """ | ||
|
|
||
| import logging | ||
| import os | ||
| from typing import List, Dict, Any, Optional | ||
|
|
||
| from langchain_openai import ChatOpenAI | ||
| from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder | ||
| from langchain.agents import AgentExecutor, create_tool_calling_agent | ||
|
|
||
|
|
||
| from tools import extract_web_text | ||
|
|
||
| logger = logging.getLogger(__name__) | ||
|
|
||
|
|
||
| def _create_llm() -> ChatOpenAI: | ||
| """Create LLM instance, supporting OpenAI and MiniMax providers.""" | ||
| if os.getenv("MINIMAX_API_KEY") and not os.getenv("OPENAI_API_KEY"): | ||
| return ChatOpenAI( | ||
| model=os.getenv("MINIMAX_MODEL", "MiniMax-M2.7"), | ||
| temperature=1.0, | ||
| api_key=os.getenv("MINIMAX_API_KEY"), | ||
| base_url=os.getenv("MINIMAX_BASE_URL", "https://api.minimax.io/v1"), | ||
| ) | ||
| return ChatOpenAI(model="gpt-4o", temperature=0) | ||
|
|
||
|
|
||
| class Agent: | ||
| def __init__(self): | ||
| # Initialize your agent | ||
| self.name = "Translation Agent" | ||
|
|
||
| # Initialize Tools | ||
| self.tools = [extract_web_text] | ||
|
|
||
| # Initialize LangChain components | ||
| self.llm = _create_llm() | ||
|
|
||
| # System prompt tailored for text-to-text translation | ||
| self.system_prompt = """You are a helpful assistant whose primary objective is to help the user with language translation. | ||
|
|
||
| RULES: | ||
| - If the user provides a URL, use the 'extract_web_text' tool to get the content, then translate the extracted text. | ||
| - Detect the source language and the target language from the user's request. | ||
| - If the user specifies a target language, translate the text to that language. | ||
| - If the user provides text without specifying a target language, default to translating it to English (if it's not English) or ask for clarification if ambiguous. | ||
| - Translate the text fully and accurately. | ||
| - Preserve the original meaning and tone. | ||
| - Do NOT add explanations or commentary unless the translation requires context notes (which should be minimal). | ||
|
|
||
| RESPONSE FORMAT: | ||
| - Provide only the translated text. | ||
| """ | ||
|
|
||
| self.prompt = ChatPromptTemplate.from_messages( | ||
| [ | ||
| ("system", self.system_prompt), | ||
| ("user", "{input}"), | ||
| MessagesPlaceholder(variable_name="agent_scratchpad"), | ||
| ] | ||
| ) | ||
|
|
||
| # Create Tool Calling Agent | ||
| agent = create_tool_calling_agent(self.llm, self.tools, self.prompt) | ||
| self.agent_executor = AgentExecutor(agent=agent, tools=self.tools, verbose=True) | ||
|
|
||
| def process_message(self, message_text: str) -> str: | ||
| """ | ||
| Process the incoming message using LangChain. | ||
| """ | ||
| logger.info(f"Processing message: {message_text[:50]}...") | ||
| result = self.agent_executor.invoke({"input": message_text}) | ||
| return result["output"] |
Comment on lines
83
to
+85
| healthcheck: | ||
| test: ["CMD-SHELL", "python -c \"import requests; requests.get('http://localhost:8000/api/v1/healthcheck').raise_for_status()\" || exit 1"] | ||
| interval: 30s | ||
| timeout: 10s | ||
| retries: 3 | ||
| disable: true | ||
|
|
Comment on lines
+93
to
+129
| def _translate_with_google( | ||
| self, text: str, src_lang: str, dest_lang: str | ||
| ) -> tuple[str, str]: | ||
| """Translate text using Google Translate API directly""" | ||
| try: | ||
| # Google Translate URL | ||
| url = "https://translate.googleapis.com/translate_a/single" | ||
|
|
||
| params = { | ||
| "client": "gtx", | ||
| "sl": src_lang, | ||
| "tl": dest_lang, | ||
| "dt": "t", | ||
| "q": text, | ||
| } | ||
|
|
||
| response = self.session.get(url, params=params, timeout=10) | ||
| response.raise_for_status() | ||
|
|
||
| result = response.json() | ||
|
|
||
| # Extract translated text | ||
| translated_text = "" | ||
| if result and len(result) > 0 and result[0]: | ||
| for sentence in result[0]: | ||
| if sentence and len(sentence) > 0: | ||
| translated_text += sentence[0] | ||
|
|
||
| # Extract detected source language | ||
| detected_src = src_lang | ||
| if len(result) > 2 and result[2]: | ||
| detected_src = result[2] | ||
|
|
||
| return translated_text.strip(), detected_src | ||
|
|
||
| except Exception as e: | ||
| raise Exception(f"Translation failed: {str(e)}") |
Comment on lines
+88
to
+94
| - a2a-sdk: A2A framework for agent communication | ||
| - googletrans: Google Translate API wrapper | ||
| - beautifulsoup4: HTML parsing for web content extraction | ||
| - langdetect: Language detection library | ||
| - requests: HTTP client for web scraping | ||
| - OpenAI: For agent conversation handling | ||
|
|
Comment on lines
+44
to
+50
| # Clone the repository | ||
| git clone https://github.com/a2aproject/a2a-samples.git | ||
| cd a2a-samples/samples/python/agents/github-agent | ||
|
|
||
| # Create virtual environment | ||
| uv venv | ||
| source .venv/bin/activate # On Windows: .venv\Scripts\activate |
Comment on lines
+85
to
+91
| messages.append( | ||
| { | ||
| "role": "assistant", | ||
| "content": message.content, | ||
| "tool_calls": message.tool_calls, | ||
| } | ||
| ) |
Comment on lines
+11
to
+13
| ports: | ||
| - "5000" | ||
| tty: true |
|
|
||
| ENV PYTHONUNBUFFERED=1 | ||
|
|
||
| CMD ["python", "__main__.py", "--host", "0.0.0.0", "--port", "5000", "--mongo-url", "mongodb://agents-mongo:27017", "--db-name", "compliance-checker-a2a"] No newline at end of file |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…ties