:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
from collections import defaultdict
from collections.abc import Mapping
from typing import Any, Optional
from posthog.schema import AssistantLifecycleQuery, LifecycleQuery
from posthog.hogql_queries.utils.result_series import get_result_order, get_result_series, get_series_key
from .utils import format_matrix, format_number, strip_datetime_seconds
STATUS_ORDER = ["new", "returning", "resurrecting", "dormant"]
STATUS_LABELS = ["New", "Returning", "Resurrecting", "Dormant"]
class LifecycleResultsFormatter:
"""
Compresses and formats lifecycle results into a LLM-friendly string.
Single series:
```
Date|New|Returning|Resurrecting|Dormant
2025-10-01|6936|29541|13263|-16735
2025-11-01|7101|30794|12662|-18946
```
Multiple series:
```
Event: $pageview
Date|New|Returning|Resurrecting|Dormant
2025-10-01|6936|29541|13263|-16735
Event: sign up
Date|New|Returning|Resurrecting|Dormant
2025-10-01|100|200|50|-80
```
"""
def __init__(
self,
query: AssistantLifecycleQuery | LifecycleQuery,
results: list[dict[str, Any]],
action_names: Optional[Mapping[int, str | None]] = None,
):
self._query = query
self._query_dict = query.model_dump(mode="json")
self._results = results
self._action_names = action_names
def format(self) -> str:
if not self._results:
return "No data recorded for this time period."
# Group results by event series
series_groups: dict[int, dict[str, dict[str, Any]]] = defaultdict(dict)
for result in self._results:
status = result.get("status", "")
series_groups[get_result_order(result)][status] = result
sections: list[str] = []
multi_series = len(series_groups) > 1
for order in sorted(series_groups.keys()):
statuses = series_groups[order]
section = self._format_series(statuses, multi_series)
if section:
sections.append(section)
return "\n\n".join(sections)
def _format_series(self, statuses: dict[str, dict[str, Any]], multi_series: bool) -> str:
# Get dates from any available status
any_result = next(iter(statuses.values()))
days = any_result.get("days", [])
if not days:
return ""
series_name = self._extract_series_name(any_result)
matrix: list[list[str]] = []
header = ["Date", *STATUS_LABELS]
matrix.append(header)
for i, day in enumerate(days):
row = [strip_datetime_seconds(day)]
for status in STATUS_ORDER:
status_result = statuses.get(status)
if status_result and i < len(status_result.get("data", [])):
row.append(format_number(status_result["data"][i]))
else:
row.append("0")
matrix.append(row)
formatted = format_matrix(matrix)
if multi_series:
return f"Event: {series_name}\n{formatted}"
return formatted
def _extract_series_name(self, result: dict[str, Any]) -> str:
node = get_result_series(self._query_dict, result)
if node is not None:
name = node.get("custom_name") or node.get("name") or get_series_key(node, self._action_names)
if name:
return name
# Fallback: strip status suffix from label
label = result.get("label", "")
for status in STATUS_ORDER:
suffix = f" - {status}"
if label.endswith(suffix):
return label[: -len(suffix)]
return label