Upload folder using huggingface_hub

POST.md

@@ -0,0 +1,35 @@
+# 30-in-15 · #12 — CSV → Insights Analyst (build-in-public post)
+
+**Attach:** a screen-recording / screenshot of a CSV going in → charts + findings coming out.
+**Demo:** https://insights.gritai.solutions · **Repo:** https://github.com/GritAI-Labs/csv-insights
+
+---
+
+📊 Project #12 of my "30 AI Projects in 15 Days": upload a CSV, get charts and a plain-English briefing back.
+
+The twist is what it *won't* do. Most "AI analyzes your data" tools let the model read the numbers and write about them — which means the model can also make numbers up. This one can't. A deterministic pandas pass computes every statistic and every chart first; the AI only gets that finished profile and turns it into prose. If a figure is in the writeup, pandas computed it. The model literally never touches the arithmetic.
+
+I went a step further on honesty. Small models love to bolt on units the data never had ("16.4°C", "3 mm") and invent methodology ("three standard deviations from the mean" when I used a different method). So there's a deterministic pass that strips guessed units, and the model is told to describe what the numbers show, not how they were computed. Percentages survive — because we compute those.
+
+What you get: shape, missing-data and outlier flags, correlations, distributions, a trend line if there's a date column, top categories — and a short briefing that leads with what's interesting and ends with the questions the data could answer next.
+
+Two tiers, same as the rest: this hosted demo writes the narrative with a hosted model; the GritAI Studio version runs the whole thing on your own local GPU fleet — free per-analysis, fully private, nothing leaves your network.
+
+▶ https://insights.gritai.solutions
+⭐ https://github.com/GritAI-Labs/csv-insights
+
+---
+
+## X / Twitter version (≤280)
+
+📊 #12 of my 30 AI projects in 15 days: upload a CSV → charts + a plain-English briefing.
+
+The catch: the AI never touches the numbers. pandas computes every stat + chart; the model only writes the story. Guessed units get stripped. Grounded by construction.
+
+→ https://github.com/GritAI-Labs/csv-insights
+
+---
+
+*(Post to LinkedIn + X. After posting, append the permalinks below.)*
+
+**Permalinks:** X: _____ · LinkedIn: _____

profiler.py

@@ -39,14 +39,16 @@ def _detect_datetime(s: pd.Series) -> pd.Series | None:
     if s.dtype == object:
         name = str(s.name).lower()
         looks_datey = any(k in name for k in ("date", "time", "day", "month", "year", "timestamp"))
-        sample = s.dropna().head(50)
+        sample = s.dropna().astype(str).head(50)
         if sample.empty:
             return None
         with warnings.catch_warnings():
             warnings.simplefilter("ignore")   # dateutil "could not infer format" is expected here
-            parsed = pd.to_datetime(sample, errors="coerce")
+            # format="mixed" parses each value on its own terms — robust across pandas versions,
+            # where bare inference could coerce valid ISO dates to NaT and mis-type the column.
+            parsed = pd.to_datetime(sample, errors="coerce", format="mixed")
             if parsed.notna().mean() >= (0.6 if looks_datey else 0.95):
-                return pd.to_datetime(s, errors="coerce")
+                return pd.to_datetime(s, errors="coerce", format="mixed")
     return None
 
 

requirements.txt

@@ -1,5 +1,9 @@
-flask>=3.0
-pandas>=2.0
-matplotlib>=3.7
+# Pinned to the exact versions verified locally (date detection + all 7 charts working).
+# Unpinned (>=) let HF's build resolve different versions where date parsing behaved differently
+# (date column mis-typed as categorical → trend chart dropped). Pin = reproduce the tested env.
+flask==3.1.3
+pandas==2.3.3
+numpy==2.2.6
+matplotlib==3.10.9
 # profiler/charts/narrator use only the above + stdlib (urllib for the local Ollama backend).
 # Optional Claude backend needs no extra package (raw HTTPS via urllib).