agentx-python 0.6.2__tar.gz → 0.6.4__tar.gz

{agentx_python-0.6.2/agentx_python.egg-info → agentx_python-0.6.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentx-python
-Version: 0.6.2
+Version: 0.6.4
 Summary: Official Python SDK for AgentX (https://www.agentx.so/)
 Home-page: https://github.com/AgentX-ai/AgentX-python
 Author: Robin Wang and AgentX Team
@@ -49,8 +49,12 @@ Dynamic: summary
 ![Logo](https://agentx-resources.s3.us-west-1.amazonaws.com/AgentX-logo-387x60.png)
 
 [![PyPI version](https://img.shields.io/pypi/v/agentx-python)](https://pypi.org/project/agentx-python/)
+[![Python versions](https://img.shields.io/pypi/pyversions/agentx-python)](https://pypi.org/project/agentx-python/)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 
-The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat with, and orchestrate AI agents in a few lines of code.
+The official Python SDK for **[AgentX](https://app.agentx.so/)** — build, chat with, orchestrate, and trace AI agents in a few lines of code.
+
+Also see [SDK Developer Docs](https://developers.agentx.so), [API Reference Docs](https://docs.agentx.so/reference)
 
 ---
 
@@ -64,8 +68,9 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
   - [List agents](#list-agents)
   - [Start a conversation](#start-a-conversation)
   - [Chat (streaming and non-streaming)](#chat-streaming-and-non-streaming)
-- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration)
-- [Agent Evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
+- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration) — teams of agents with a designated manager
+- [Production tracing](#production-tracing) — record live agent runs from any framework
+- [Custom agent evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
 - [Links](#links)
 
 ---
@@ -78,6 +83,7 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
 - **Batteries included** — voice (ASR/TTS), image generation, document/CSV/Excel/OCR, RAG with built-in re-ranking.
 - **MCP support** — connect any Model Context Protocol server.
 - **Multi-agent orchestration** — workforces of agents with a designated manager, across LLM vendors.
+- **Production tracing** — one decorator or context manager records every agent run (input, output, latency, tool calls, token usage) into your workspace, for any framework.
 - **Agent Evaluations** — score any agent (LangChain, CrewAI, OpenAI, Anthropic, HTTP, …) with LLM-as-a-judge ratings plus optional cosine and Jaccard similarity metrics.
 - **A2A** — Each agent can be published with agent-to-agent protocol compatible.
 
@@ -187,6 +193,58 @@ for chunk in workforce.chat_stream(conversation.id, "How can you help me with th
 
 ---
 
+## Production tracing
+
+Record live agent runs into your workspace with a single decorator or context manager — no changes to your agent's logic. Traces appear in the **Live Traces** tab and can be evaluated against your test datasets with [`tracer.evaluate_trace()`](TRACING.md#tracerevaluate_trace).
+
+```python
+from agentx import AgentX
+
+client = AgentX.from_env()
+tracer = client.tracer
+
+@tracer.trace("customer-support-agent", framework="langchain", model="gpt-4o")
+def handle_query(query: str) -> str:
+    return chain.invoke(query)
+
+# Every call is automatically traced: input, output, latency, tool calls, token usage
+handle_query("How do I reset my password?")
+tracer.flush(timeout=10)  # ensure delivery before the process exits
+```
+
+Prefer full control over what gets captured? Use the context manager instead:
+
+```python
+with tracer.trace("rag-agent", framework="langchain") as span:
+    span.input = {"query": query, "user_id": user_id}
+
+    kb_result = search_knowledge_base(query)
+    span.add_tool_call("search_knowledge_base", input=query, output=kb_result, latency_ms=190)
+
+    span.output = llm.invoke(f"Context: {kb_result}\n\nQuery: {query}")
+```
+
+### Framework integrations
+
+Each integration auto-captures LLM calls, tool calls, and token usage — install the matching extra:
+
+| Framework             | Install                                      | Integration              |
+| --------------------- | -------------------------------------------- | ------------------------ |
+| LangChain             | `pip install "agentx-python[langchain]"`     | `AgentXCallbackHandler`  |
+| CrewAI                | `pip install "agentx-python[crewai]"`        | `AgentXCrewObserver`     |
+| OpenAI Agents SDK     | `pip install "agentx-python[openai-agents]"` | `AgentXTracingProcessor` |
+| Anthropic             | `pip install "agentx-python[anthropic]"`     | `patch_anthropic_client` |
+| Google ADK            | `pip install "agentx-python[google-adk]"`    | `AgentXADKPlugin`        |
+| Google GenAI (Gemini) | `pip install "agentx-python[google-genai]"`  | `patch_genai_client`     |
+
+Or plain Python — wrap any function with `@tracer.trace(...)` and it just works, no framework required.
+
+Running specialist agents in parallel with a `ThreadPoolExecutor`? Wrap each worker body in `tracer.use_span(span)` so their steps land on the parent trace instead of becoming independent traces — see [TRACING.md](TRACING.md) for the full pattern.
+
+See **[TRACING.md](TRACING.md)** for the complete guide — session grouping, error handling, async support, and the full API reference.
+
+---
+
 ## Custom agent evaluations
 
 Evaluate **any** AI agent — LangChain, CrewAI, AutoGen, LlamaIndex, OpenAI, Anthropic, HTTP endpoints, or plain Python — using AgentX as the scoring and reporting backend. Includes optional **cosine** and **Jaccard** similarity metrics alongside LLM-graded ratings.
@@ -214,4 +272,5 @@ See **[EVALUATIONS.md](EVALUATIONS.md)** for the full guide — dataset builder,
 - **Dashboard** — [app.agentx.so](https://app.agentx.so)
 - **Website** — [agentx.so](https://www.agentx.so/)
 - **PyPI** — [agentx-python](https://pypi.org/project/agentx-python/)
+- **Tracing docs** — [TRACING.md](TRACING.md)
 - **Evaluations docs** — [EVALUATIONS.md](EVALUATIONS.md)

{agentx_python-0.6.2 → agentx_python-0.6.4}/README.md

@@ -1,8 +1,12 @@
 ![Logo](https://agentx-resources.s3.us-west-1.amazonaws.com/AgentX-logo-387x60.png)
 
 [![PyPI version](https://img.shields.io/pypi/v/agentx-python)](https://pypi.org/project/agentx-python/)
+[![Python versions](https://img.shields.io/pypi/pyversions/agentx-python)](https://pypi.org/project/agentx-python/)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 
-The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat with, and orchestrate AI agents in a few lines of code.
+The official Python SDK for **[AgentX](https://app.agentx.so/)** — build, chat with, orchestrate, and trace AI agents in a few lines of code.
+
+Also see [SDK Developer Docs](https://developers.agentx.so), [API Reference Docs](https://docs.agentx.so/reference)
 
 ---
 
@@ -16,8 +20,9 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
   - [List agents](#list-agents)
   - [Start a conversation](#start-a-conversation)
   - [Chat (streaming and non-streaming)](#chat-streaming-and-non-streaming)
-- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration)
-- [Agent Evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
+- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration) — teams of agents with a designated manager
+- [Production tracing](#production-tracing) — record live agent runs from any framework
+- [Custom agent evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
 - [Links](#links)
 
 ---
@@ -30,6 +35,7 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
 - **Batteries included** — voice (ASR/TTS), image generation, document/CSV/Excel/OCR, RAG with built-in re-ranking.
 - **MCP support** — connect any Model Context Protocol server.
 - **Multi-agent orchestration** — workforces of agents with a designated manager, across LLM vendors.
+- **Production tracing** — one decorator or context manager records every agent run (input, output, latency, tool calls, token usage) into your workspace, for any framework.
 - **Agent Evaluations** — score any agent (LangChain, CrewAI, OpenAI, Anthropic, HTTP, …) with LLM-as-a-judge ratings plus optional cosine and Jaccard similarity metrics.
 - **A2A** — Each agent can be published with agent-to-agent protocol compatible.
 
@@ -139,6 +145,58 @@ for chunk in workforce.chat_stream(conversation.id, "How can you help me with th
 
 ---
 
+## Production tracing
+
+Record live agent runs into your workspace with a single decorator or context manager — no changes to your agent's logic. Traces appear in the **Live Traces** tab and can be evaluated against your test datasets with [`tracer.evaluate_trace()`](TRACING.md#tracerevaluate_trace).
+
+```python
+from agentx import AgentX
+
+client = AgentX.from_env()
+tracer = client.tracer
+
+@tracer.trace("customer-support-agent", framework="langchain", model="gpt-4o")
+def handle_query(query: str) -> str:
+    return chain.invoke(query)
+
+# Every call is automatically traced: input, output, latency, tool calls, token usage
+handle_query("How do I reset my password?")
+tracer.flush(timeout=10)  # ensure delivery before the process exits
+```
+
+Prefer full control over what gets captured? Use the context manager instead:
+
+```python
+with tracer.trace("rag-agent", framework="langchain") as span:
+    span.input = {"query": query, "user_id": user_id}
+
+    kb_result = search_knowledge_base(query)
+    span.add_tool_call("search_knowledge_base", input=query, output=kb_result, latency_ms=190)
+
+    span.output = llm.invoke(f"Context: {kb_result}\n\nQuery: {query}")
+```
+
+### Framework integrations
+
+Each integration auto-captures LLM calls, tool calls, and token usage — install the matching extra:
+
+| Framework             | Install                                      | Integration              |
+| --------------------- | -------------------------------------------- | ------------------------ |
+| LangChain             | `pip install "agentx-python[langchain]"`     | `AgentXCallbackHandler`  |
+| CrewAI                | `pip install "agentx-python[crewai]"`        | `AgentXCrewObserver`     |
+| OpenAI Agents SDK     | `pip install "agentx-python[openai-agents]"` | `AgentXTracingProcessor` |
+| Anthropic             | `pip install "agentx-python[anthropic]"`     | `patch_anthropic_client` |
+| Google ADK            | `pip install "agentx-python[google-adk]"`    | `AgentXADKPlugin`        |
+| Google GenAI (Gemini) | `pip install "agentx-python[google-genai]"`  | `patch_genai_client`     |
+
+Or plain Python — wrap any function with `@tracer.trace(...)` and it just works, no framework required.
+
+Running specialist agents in parallel with a `ThreadPoolExecutor`? Wrap each worker body in `tracer.use_span(span)` so their steps land on the parent trace instead of becoming independent traces — see [TRACING.md](TRACING.md) for the full pattern.
+
+See **[TRACING.md](TRACING.md)** for the complete guide — session grouping, error handling, async support, and the full API reference.
+
+---
+
 ## Custom agent evaluations
 
 Evaluate **any** AI agent — LangChain, CrewAI, AutoGen, LlamaIndex, OpenAI, Anthropic, HTTP endpoints, or plain Python — using AgentX as the scoring and reporting backend. Includes optional **cosine** and **Jaccard** similarity metrics alongside LLM-graded ratings.
@@ -166,4 +224,5 @@ See **[EVALUATIONS.md](EVALUATIONS.md)** for the full guide — dataset builder,
 - **Dashboard** — [app.agentx.so](https://app.agentx.so)
 - **Website** — [agentx.so](https://www.agentx.so/)
 - **PyPI** — [agentx-python](https://pypi.org/project/agentx-python/)
+- **Tracing docs** — [TRACING.md](TRACING.md)
 - **Evaluations docs** — [EVALUATIONS.md](EVALUATIONS.md)

{agentx_python-0.6.2 → agentx_python-0.6.4}/agentx/integrations/anthropic.py

@@ -22,6 +22,34 @@ from agentx.tracing.tracer import Tracer, _safe_serialize
 from agentx.integrations._perf import build_performance_summary
 
 
+def _extract_output_text(response: Any) -> Optional[str]:
+    """
+    Extract the assistant's text reply from a Messages API response, falling
+    back to a description of any tool_use blocks when the response is a pure
+    tool call with no accompanying text.
+    """
+    content = getattr(response, "content", None) if response is not None else None
+    if not content:
+        return None
+    texts = []
+    tool_calls = []
+    for block in content:
+        block_type = getattr(block, "type", None)
+        if block_type == "text":
+            text = getattr(block, "text", None)
+            if text:
+                texts.append(text)
+        elif block_type == "tool_use":
+            name = getattr(block, "name", "unknown")
+            tool_input = getattr(block, "input", None)
+            tool_calls.append(f"{name}({tool_input})")
+    if texts:
+        return "\n".join(texts)
+    if tool_calls:
+        return "[tool call] " + ", ".join(tool_calls)
+    return None
+
+
 def patch_anthropic_client(
     client: Any,
     tracer: Tracer,
@@ -77,10 +105,7 @@ def _patch_create(
             input_tokens = None
             output_tokens = None
             if response is not None:
-                try:
-                    output = response.content[0].text if response.content else None
-                except Exception:
-                    output = str(response)
+                output = _extract_output_text(response)
                 try:
                     usage = getattr(response, "usage", None)
                     if usage is not None:
@@ -172,7 +197,7 @@ def _patch_stream(
                 output_tokens = None
                 try:
                     final = ctx.get_final_message()
-                    output = final.content[0].text if final.content else None
+                    output = _extract_output_text(final)
                     usage = getattr(final, "usage", None)
                     if usage is not None:
                         input_tokens = getattr(usage, "input_tokens", None)

{agentx_python-0.6.2 → agentx_python-0.6.4}/agentx/integrations/google_adk.py

@@ -36,7 +36,11 @@ except ImportError as exc:  # pragma: no cover
 
 
 def _content_to_text(content: Any) -> Optional[str]:
-    """Extract plain text from a google.genai types.Content object."""
+    """
+    Extract plain text from a google.genai types.Content object, falling back
+    to a description of any function_call parts when there's no text (Gemini
+    function calling — the model responded with a pure tool call).
+    """
     if content is None:
         return None
     if isinstance(content, str):
@@ -45,11 +49,22 @@ def _content_to_text(content: Any) -> Optional[str]:
     if not parts:
         return None
     texts = []
+    function_calls = []
     for part in parts:
         text = getattr(part, "text", None)
         if text and isinstance(text, str):
             texts.append(text)
-    return " ".join(texts) if texts else None
+            continue
+        fc = getattr(part, "function_call", None)
+        if fc is not None:
+            name = getattr(fc, "name", "unknown")
+            args = getattr(fc, "args", None)
+            function_calls.append(f"{name}({args})")
+    if texts:
+        return " ".join(texts)
+    if function_calls:
+        return "[tool call] " + ", ".join(function_calls)
+    return None
 
 
 def _contents_to_text(contents: Any) -> Optional[str]:

{agentx_python-0.6.2 → agentx_python-0.6.4}/agentx/integrations/google_genai.py

@@ -16,7 +16,7 @@ Requires: ``pip install "agentx-python[google-genai]"``
 from __future__ import annotations
 
 import time
-from typing import Any, Dict, Optional
+from typing import Any, Dict, List, Optional
 
 from agentx.tracing.tracer import Tracer, _safe_serialize
 from agentx.integrations._perf import build_performance_summary
@@ -49,13 +49,18 @@ def patch_genai_client(
 
 
 def _extract_response_text(response: Any) -> Optional[str]:
-    """Pull the generated text out of a GenerateContentResponse."""
+    """
+    Pull the generated text out of a GenerateContentResponse, falling back to
+    a description of any function_call parts when the response is a pure
+    tool call with no text (Gemini function calling).
+    """
     # Convenience .text property (available on non-streaming responses)
     text = getattr(response, "text", None)
     if text and isinstance(text, str):
         return text
     # Fallback: walk candidates → content → parts
     candidates = getattr(response, "candidates", None) or []
+    function_calls: List[str] = []
     for candidate in candidates:
         content = getattr(candidate, "content", None)
         parts = getattr(content, "parts", None) or []
@@ -63,6 +68,13 @@ def _extract_response_text(response: Any) -> Optional[str]:
             t = getattr(part, "text", None)
             if t and isinstance(t, str):
                 return t
+            fc = getattr(part, "function_call", None)
+            if fc is not None:
+                name = getattr(fc, "name", "unknown")
+                args = getattr(fc, "args", None)
+                function_calls.append(f"{name}({args})")
+    if function_calls:
+        return "[tool call] " + ", ".join(function_calls)
     return None
 
 

{agentx_python-0.6.2 → agentx_python-0.6.4}/agentx/integrations/langchain.py

@@ -177,9 +177,30 @@ def _extract_llm_input(
     return None
 
 
+def _describe_tool_calls(message: Any) -> Optional[str]:
+    """
+    Format an AIMessage's ``tool_calls`` as a readable fallback for ``output``
+    when the model responded with a pure tool call and no text content.
+    """
+    tool_calls = getattr(message, "tool_calls", None) if message is not None else None
+    if not tool_calls:
+        return None
+    parts = []
+    for tc in tool_calls:
+        if isinstance(tc, dict):
+            name = tc.get("name") or "unknown"
+            args = tc.get("args")
+        else:
+            name = getattr(tc, "name", None) or "unknown"
+            args = getattr(tc, "args", None)
+        parts.append(f"{name}({args})" if args is not None else f"{name}()")
+    return "[tool call] " + ", ".join(parts)
+
+
 def _extract_llm_output(response: "LLMResult") -> Optional[str]:
     """Flatten on_llm_end's ``LLMResult`` (chat or completion generations) into one string."""
     texts: List[str] = []
+    tool_call_fallbacks: List[str] = []
     for gen_list in getattr(response, "generations", None) or []:
         for gen in gen_list or []:
             message = getattr(gen, "message", None)
@@ -196,7 +217,18 @@ def _extract_llm_output(response: "LLMResult") -> Optional[str]:
             text = getattr(gen, "text", None)
             if text:
                 texts.append(text)
-    return "\n".join(texts) if texts else None
+                continue
+            # No text content — the model likely responded with a pure tool
+            # call instead of commentary. Fall back to describing it so the
+            # step's output isn't silently omitted.
+            described = _describe_tool_calls(message)
+            if described:
+                tool_call_fallbacks.append(described)
+    if texts:
+        return "\n".join(texts)
+    if tool_call_fallbacks:
+        return "\n".join(tool_call_fallbacks)
+    return None
 
 
 class AgentXCallbackHandler(BaseCallbackHandler):

{agentx_python-0.6.2 → agentx_python-0.6.4}/agentx/tracing/tracer.py

@@ -4,6 +4,7 @@ import asyncio
 import concurrent.futures
 import functools
 import inspect
+import re
 import threading
 import time
 from contextlib import contextmanager
@@ -20,6 +21,11 @@ from agentx.integrations._perf import (
 
 F = TypeVar("F", bound=Callable[..., Any])
 
+# Matches the generic "LLM Call N" step names integrations generate (e.g.
+# langchain.py numbers steps within one top-level chain run). Used by
+# _TraceSpan._merge_child_run to renumber them on merge — see there.
+_LLM_CALL_NAME_RE = re.compile(r"^LLM Call \d+$")
+
 
 def _safe_serialize(value: Any, depth: int = 0) -> Any:
     """Best-effort conversion to a JSON-safe structure, truncated to avoid huge payloads."""
@@ -208,6 +214,19 @@ class _TraceSpan:
         becoming its own independent trace.
         """
         with self._merge_lock:
+            if execution_steps:
+                # Step names like "LLM Call N" are numbered locally within
+                # whatever sub-run produced them (e.g. one LangChain chain
+                # invocation numbers its own calls 1, 2, 3...). When several
+                # sub-runs merge into this span — e.g. a sequence of
+                # single-call specialist chains all folded into one
+                # orchestrator span via use_span() — renumber so the merged
+                # trace doesn't end up with several "LLM Call 1" entries.
+                next_n = len(self._execution_steps) + 1
+                for offset, step in enumerate(execution_steps):
+                    name = step.get("name")
+                    if isinstance(name, str) and _LLM_CALL_NAME_RE.match(name):
+                        step["name"] = f"LLM Call {next_n + offset}"
             self._execution_steps.extend(execution_steps or [])
             self.tool_calls.extend(tool_calls or [])
             self._retrieval_steps.extend(retrieval_steps or [])

agentx_python-0.6.4/agentx/version.py

@@ -0,0 +1 @@
+VERSION = "0.6.4"

{agentx_python-0.6.2 → agentx_python-0.6.4/agentx_python.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentx-python
-Version: 0.6.2
+Version: 0.6.4
 Summary: Official Python SDK for AgentX (https://www.agentx.so/)
 Home-page: https://github.com/AgentX-ai/AgentX-python
 Author: Robin Wang and AgentX Team
@@ -49,8 +49,12 @@ Dynamic: summary
 ![Logo](https://agentx-resources.s3.us-west-1.amazonaws.com/AgentX-logo-387x60.png)
 
 [![PyPI version](https://img.shields.io/pypi/v/agentx-python)](https://pypi.org/project/agentx-python/)
+[![Python versions](https://img.shields.io/pypi/pyversions/agentx-python)](https://pypi.org/project/agentx-python/)
+[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 
-The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat with, and orchestrate AI agents in a few lines of code.
+The official Python SDK for **[AgentX](https://app.agentx.so/)** — build, chat with, orchestrate, and trace AI agents in a few lines of code.
+
+Also see [SDK Developer Docs](https://developers.agentx.so), [API Reference Docs](https://docs.agentx.so/reference)
 
 ---
 
@@ -64,8 +68,9 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
   - [List agents](#list-agents)
   - [Start a conversation](#start-a-conversation)
   - [Chat (streaming and non-streaming)](#chat-streaming-and-non-streaming)
-- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration)
-- [Agent Evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
+- [Workforce (multi-agent orchestration)](#workforce-multi-agent-orchestration) — teams of agents with a designated manager
+- [Production tracing](#production-tracing) — record live agent runs from any framework
+- [Custom agent evaluations](#custom-agent-evaluations) — LLM-as-a-judge, cosine / Jaccard similarity
 - [Links](#links)
 
 ---
@@ -78,6 +83,7 @@ The official Python SDK for **[AgentX](https://www.agentx.so/)** — build, chat
 - **Batteries included** — voice (ASR/TTS), image generation, document/CSV/Excel/OCR, RAG with built-in re-ranking.
 - **MCP support** — connect any Model Context Protocol server.
 - **Multi-agent orchestration** — workforces of agents with a designated manager, across LLM vendors.
+- **Production tracing** — one decorator or context manager records every agent run (input, output, latency, tool calls, token usage) into your workspace, for any framework.
 - **Agent Evaluations** — score any agent (LangChain, CrewAI, OpenAI, Anthropic, HTTP, …) with LLM-as-a-judge ratings plus optional cosine and Jaccard similarity metrics.
 - **A2A** — Each agent can be published with agent-to-agent protocol compatible.
 
@@ -187,6 +193,58 @@ for chunk in workforce.chat_stream(conversation.id, "How can you help me with th
 
 ---
 
+## Production tracing
+
+Record live agent runs into your workspace with a single decorator or context manager — no changes to your agent's logic. Traces appear in the **Live Traces** tab and can be evaluated against your test datasets with [`tracer.evaluate_trace()`](TRACING.md#tracerevaluate_trace).
+
+```python
+from agentx import AgentX
+
+client = AgentX.from_env()
+tracer = client.tracer
+
+@tracer.trace("customer-support-agent", framework="langchain", model="gpt-4o")
+def handle_query(query: str) -> str:
+    return chain.invoke(query)
+
+# Every call is automatically traced: input, output, latency, tool calls, token usage
+handle_query("How do I reset my password?")
+tracer.flush(timeout=10)  # ensure delivery before the process exits
+```
+
+Prefer full control over what gets captured? Use the context manager instead:
+
+```python
+with tracer.trace("rag-agent", framework="langchain") as span:
+    span.input = {"query": query, "user_id": user_id}
+
+    kb_result = search_knowledge_base(query)
+    span.add_tool_call("search_knowledge_base", input=query, output=kb_result, latency_ms=190)
+
+    span.output = llm.invoke(f"Context: {kb_result}\n\nQuery: {query}")
+```
+
+### Framework integrations
+
+Each integration auto-captures LLM calls, tool calls, and token usage — install the matching extra:
+
+| Framework             | Install                                      | Integration              |
+| --------------------- | -------------------------------------------- | ------------------------ |
+| LangChain             | `pip install "agentx-python[langchain]"`     | `AgentXCallbackHandler`  |
+| CrewAI                | `pip install "agentx-python[crewai]"`        | `AgentXCrewObserver`     |
+| OpenAI Agents SDK     | `pip install "agentx-python[openai-agents]"` | `AgentXTracingProcessor` |
+| Anthropic             | `pip install "agentx-python[anthropic]"`     | `patch_anthropic_client` |
+| Google ADK            | `pip install "agentx-python[google-adk]"`    | `AgentXADKPlugin`        |
+| Google GenAI (Gemini) | `pip install "agentx-python[google-genai]"`  | `patch_genai_client`     |
+
+Or plain Python — wrap any function with `@tracer.trace(...)` and it just works, no framework required.
+
+Running specialist agents in parallel with a `ThreadPoolExecutor`? Wrap each worker body in `tracer.use_span(span)` so their steps land on the parent trace instead of becoming independent traces — see [TRACING.md](TRACING.md) for the full pattern.
+
+See **[TRACING.md](TRACING.md)** for the complete guide — session grouping, error handling, async support, and the full API reference.
+
+---
+
 ## Custom agent evaluations
 
 Evaluate **any** AI agent — LangChain, CrewAI, AutoGen, LlamaIndex, OpenAI, Anthropic, HTTP endpoints, or plain Python — using AgentX as the scoring and reporting backend. Includes optional **cosine** and **Jaccard** similarity metrics alongside LLM-graded ratings.
@@ -214,4 +272,5 @@ See **[EVALUATIONS.md](EVALUATIONS.md)** for the full guide — dataset builder,
 - **Dashboard** — [app.agentx.so](https://app.agentx.so)
 - **Website** — [agentx.so](https://www.agentx.so/)
 - **PyPI** — [agentx-python](https://pypi.org/project/agentx-python/)
+- **Tracing docs** — [TRACING.md](TRACING.md)
 - **Evaluations docs** — [EVALUATIONS.md](EVALUATIONS.md)

agentx_python-0.6.2/agentx/version.py

@@ -1 +0,0 @@
-VERSION = "0.6.2"