From 8e9906d844ee4a913cbd9ac8c162f8fdc7f06ac7 Mon Sep 17 00:00:00 2001
From: Inferencebot <bot@inference.coop>
Date: Wed, 28 Oct 2026 04:33:50 +0000
Subject: [PATCH] =?UTF-8?q?metrics:=20fix=20the=20impact=20columns=20?=
 =?UTF-8?q?=E2=80=94=20group=20by=20model=5Fgroup,=20never=20write=20a=20z?=
 =?UTF-8?q?ero?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Two bugs from the first deploy, both caught by reading the numbers back:

1. Tokens were summed per raw `model`, which carries the UPSTREAM name
   (openai/deepseek-v4.1-flash), while the factor table is keyed on our
   provider aliases (greenpt/deepseek-v4.1-flash). Every model therefore
   matched nothing. Now grouped on model_group, the same convention
   aggregate_model_usage uses, with model as the fallback for old rows.

2. Every day was written as energy_wh = 0.0 / co2e_g = 0.0 — a false claim that
   the co-op used no energy, and precisely the fail-open this module is written
   to avoid. A day with no measured tokens now yields NULL for all three impact
   columns, and impact_for() refuses to compute from an empty total.

The backfill is idempotent, so re-running it overwrote the zero rows.
---
 metrics_snapshot.py | 25 ++++++++++++++++++-------
 1 file changed, 18 insertions(+), 7 deletions(-)

diff --git a/metrics_snapshot.py b/metrics_snapshot.py
index 20122d7..6c2a9b1 100644
--- a/metrics_snapshot.py
+++ b/metrics_snapshot.py
@@ -294,16 +294,21 @@ def tokens_by_day_model(since: dt.date | None = None) -> dict[str, dict[str, int
         if since:
             where, params = 'WHERE "startTime" >= %s', [since.isoformat()]
         with conn.cursor() as cur:
+            # model_group carries our provider/model alias (e.g.
+            # greenpt/deepseek-v4.1-flash) — the name the factor table is keyed
+            # on. Raw `model` holds the upstream name (openai/deepseek-v4.1-flash)
+            # and would match nothing. Same convention as aggregate_model_usage.
             cur.execute(
-                f"""SELECT to_char("startTime", 'YYYY-MM-DD') AS day, model,
+                f"""SELECT to_char("startTime", 'YYYY-MM-DD') AS day,
+                           COALESCE(NULLIF(model_group, ''), NULLIF(model, ''), 'unknown') AS alias,
                            COALESCE(SUM(completion_tokens), 0)
                     FROM "LiteLLM_SpendLogs" {where}
                     GROUP BY 1, 2""",
                 params,
             )
             out: dict[str, dict[str, int]] = {}
-            for day, model, toks in cur.fetchall():
-                out.setdefault(day, {})[model or "unknown"] = int(toks)
+            for day, alias, toks in cur.fetchall():
+                out.setdefault(day, {})[alias or "unknown"] = int(toks)
     finally:
         conn.close()
     return out
@@ -319,6 +324,12 @@ def impact_for(models: dict[str, int]) -> tuple[float | None, float | None]:
     """
     if not models:
         return None, None
+    if not sum(t for t in models.values() if t):
+        # No measured tokens for this day: report "not measured" (NULL), never a
+        # zero. A zero reads as "the co-op used no energy" — a lie that would
+        # survive into a chart. The first backfill shipped exactly that, because
+        # it summed the wrong column and every model totalled 0.
+        return None, None
     wh = 0.0
     co2 = 0.0
     for model, toks in models.items():
@@ -388,9 +399,9 @@ def snapshot(day: str | None = None) -> dict:
     spend = spend_by_day(since=since)
     today_spend = spend.get(day, (None, None))
     day_models = (tokens_by_day_model(since=since).get(day)) or {}
-    out_tokens, energy_wh, co2e_g = (
-        (sum(day_models.values()), *impact_for(day_models)) if day_models else (None, None, None)
-    )
+    tot = sum(day_models.values())
+    out_tokens = tot or None
+    energy_wh, co2e_g = impact_for(day_models) if tot else (None, None)
     row = {
         "members_active": m["active"],
         "members_total": m["total"],
@@ -449,7 +460,7 @@ def backfill() -> None:
             "requests": spend.get(key, (None, None))[1],
         }
         day_models = tokens.get(key) or {}
-        if day_models:
+        if sum(day_models.values()):
             row["out_tokens"] = sum(day_models.values())
             row["energy_wh"], row["co2e_g"] = impact_for(day_models)
         if key == today():