Add PostHog analytics with graceful degradation

Server-side tracking (started/completed/failed) and client-side JS snippet,
gated behind _POSTHOG_ENABLED so calls are skipped when env vars are unset.
API keys are hashed for distinct_id. Also cleans up unused _extract_answer
and RESEARCH_TRUNCATE_LEN.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app.py

@@ -1,4 +1,5 @@
 import ast
+import hashlib
 import json
 import os
 import tempfile
@@ -6,8 +7,23 @@ from typing import Literal
 
 import gradio as gr
 import pandas as pd
+import posthog
 from pydantic import BaseModel, Field, create_model
 
+_POSTHOG_KEY = os.environ.get("POSTHOG_KEY", "")
+_POSTHOG_HOST = os.environ.get("POSTHOG_HOST", "")
+_POSTHOG_ENABLED = bool(_POSTHOG_KEY and _POSTHOG_HOST)
+
+posthog.project_api_key = _POSTHOG_KEY
+posthog.host = _POSTHOG_HOST
+
+POSTHOG_HEAD = f"""
+<script>
+  !function(t,e){{var o,n,p,r;e.__SV||(window.posthog=e,e._i=[],e.init=function(i,s,a){{function g(t,e){{var o=e.split(".");2==o.length&&(t=t[o[0]],e=o[1]),t[e]=function(){{t.push([e].concat(Array.prototype.slice.call(arguments,0)))}}}}(p=t.createElement("script")).type="text/javascript",p.async=!0,p.src=s.api_host+"/static/array.js",(r=t.getElementsByTagName("script")[0]).parentNode.insertBefore(p,r);var u=e;for(void 0!==a?u=e[a]=[]:a="posthog",u.people=u.people||[],u.toString=function(t){{var e="posthog";return"posthog"!==a&&(e+="."+a),t||(e+=" (stub)"),e}},u.people.toString=function(){{return u.toString(1)+".people (stub)"}},o="init capture register register_once unregister opt_in_capturing opt_out_capturing has_opted_in_capturing has_opted_out_capturing identify alias people.set people.set_once set_config reset opt_in_capturing".split(" "),n=0;n<o.length;n++)g(u,o[n]);e._i.push([i,s,a])}},e.__SV=1)}}(document,window.posthog||[]);
+  posthog.init('{_POSTHOG_KEY}',{{api_host:'{_POSTHOG_HOST}', person_profiles: 'identified_only'}})
+</script>
+""" if _POSTHOG_ENABLED else ""
+
 from everyrow.ops import agent_map
 from everyrow.task import EffortLevel
 
@@ -107,6 +123,11 @@ def fields_to_schema_json(fields_list: list[dict]) -> str:
     return json.dumps(clean, indent=2) if clean else ""
 
 
+def _posthog_distinct_id(api_key: str) -> str:
+    """Hash the API key to use as a stable distinct_id without storing the secret."""
+    return hashlib.sha256(api_key.encode()).hexdigest()[:16]
+
+
 async def run_agent_map(api_key, file, query, effort_label, fields_list):
     if not api_key:
         raise gr.Error("Please enter your everyrow API key.")
@@ -122,6 +143,7 @@ async def run_agent_map(api_key, file, query, effort_label, fields_list):
         raise gr.Error("The uploaded CSV is empty.")
 
     effort_level = EFFORT_LEVELS[effort_label]
+    distinct_id = _posthog_distinct_id(api_key)
 
     try:
         response_model = build_response_model(fields_list)
@@ -132,11 +154,41 @@ async def run_agent_map(api_key, file, query, effort_label, fields_list):
     if response_model is not None:
         kwargs["response_model"] = response_model
 
+    if _POSTHOG_ENABLED:
+        posthog.capture(
+            distinct_id=distinct_id,
+            event="agent_map_started",
+            properties={
+                "effort_level": effort_label,
+                "row_count": len(df),
+                "column_count": len(df.columns),
+                "field_count": len(fields_list) if fields_list else 0,
+                "app": "everyrow-research-space",
+            },
+        )
+
     result = await agent_map(**kwargs)
 
     if result.error:
+        if _POSTHOG_ENABLED:
+            posthog.capture(
+                distinct_id=distinct_id,
+                event="agent_map_failed",
+                properties={"error": str(result.error), "app": "everyrow-research-space"},
+            )
         raise gr.Error(f"agent_map failed: {result.error}")
 
+    if _POSTHOG_ENABLED:
+        posthog.capture(
+            distinct_id=distinct_id,
+            event="agent_map_completed",
+            properties={
+                "output_rows": len(result.data),
+                "output_columns": len(result.data.columns),
+                "app": "everyrow-research-space",
+            },
+        )
+
     output_df = result.data
 
     if "research" in output_df.columns:
@@ -151,9 +203,6 @@ async def run_agent_map(api_key, file, query, effort_label, fields_list):
     return output_df, gr.update(value=tmp.name, visible=True)
 
 
-RESEARCH_TRUNCATE_LEN = 120
-
-
 def _parse_research_val(val):
     """Try to parse a research value into a dict."""
     if isinstance(val, dict):
@@ -168,14 +217,6 @@ def _parse_research_val(val):
     return None
 
 
-def _extract_answer(val):
-    """Extract the answer string from a research value."""
-    parsed = _parse_research_val(val)
-    if parsed is not None:
-        return parsed.get("answer", str(val))
-    return str(val)
-
-
 def hide_research(df: pd.DataFrame) -> pd.DataFrame:
     """Default view: drop the research column."""
     if "research" not in df.columns:
@@ -203,6 +244,7 @@ def expand_research(df: pd.DataFrame) -> pd.DataFrame:
 
 with gr.Blocks(
     title="everyrow annotate – AI Data Annotation & Web Research",
+    head=POSTHOG_HEAD,
     css=".error-box { background: #fee; border: 1px solid #c00; border-radius: 8px; padding: 12px; color: #900; }",
 ) as demo:
     gr.Markdown(

requirements.txt

@@ -1 +1,2 @@
 everyrow>=0.2.0
+posthog>=3.0.0