Upload rebuilt Space (research showcase + labelled demo)

Dockerfile

@@ -16,8 +16,10 @@ RUN pip install --no-cache-dir -r requirements.txt
 # Copy ALL your files (including run.sh, streamlit_app.py, etc.) into the container
 COPY . .
 
-# NOW that run.sh is copied, make it executable
-RUN chmod +x ./run.sh
+# NOW that run.sh is copied, strip Windows line endings (a CRLF shebang makes the
+# container die with "exec ./run.sh: no such file or directory", 2026-10-31) and make
+# it executable.
+RUN sed -i 's/\r$//' ./run.sh && chmod +x ./run.sh
 
 # This is the command that will run when the Space starts
 CMD ["./run.sh"]

run.sh

@@ -1,64 +1,64 @@
-#!/bin/bash
-# Space entrypoint: FastAPI on :7860 (internal) + Streamlit on :8501 (the exposed port).
-#
-# The old version was two lines -- background uvicorn with `&`, then start Streamlit -- and
-# never checked whether the API came up. That matters because config.py raises ImportError
-# when FINNHUB_KEY or REDDIT_CLIENT_ID is unset, which kills uvicorn instantly. Streamlit
-# then starts fine, the Space reports "Running" in green, and every click returns
-# "Failed to connect to the API" with no indication that the actual cause is a missing
-# secret. Same failure shape as every other bug in this codebase: a real error absorbed by
-# the layer above and re-presented as something vague.
-#
-# So: wait for the API to actually answer, and if it does not, write the reason somewhere
-# the UI can read and display. Streamlit still starts either way -- a Space that boots and
-# explains itself is far more useful than one that exits.
-set -u
-
-STATUS_FILE=/tmp/api_status
-API_LOG=/tmp/api.log
-: > "$STATUS_FILE"
-
-echo "[run.sh] starting FastAPI on :7860 ..."
-uvicorn trade_analysis.enhanced_api:app --host 0.0.0.0 --port 7860 > "$API_LOG" 2>&1 &
-API_PID=$!
-
-# python, not curl -- python:3.12-slim ships no curl and adding it just for a health check
-# is a layer nobody needs.
-ready=0
-for _ in $(seq 1 45); do
-    if ! kill -0 "$API_PID" 2>/dev/null; then
-        {
-            echo "The analysis API exited while starting up."
-            echo
-            echo "Most common cause: a required Space secret is not set."
-            echo "Settings -> Variables and secrets: FINNHUB_KEY, REDDIT_CLIENT_ID,"
-            echo "REDDIT_CLIENT_SECRET, REDDIT_USER_AGENT."
-            echo
-            echo "--- last lines of the API log ---"
-            tail -n 25 "$API_LOG"
-        } > "$STATUS_FILE"
-        echo "[run.sh] API process died during startup; see $STATUS_FILE"
-        break
-    fi
-    if python -c "import urllib.request; urllib.request.urlopen('http://localhost:7860/', timeout=2)" 2>/dev/null; then
-        echo "ready" > "$STATUS_FILE"
-        ready=1
-        echo "[run.sh] API is answering."
-        break
-    fi
-    sleep 2
-done
-
-if [ "$ready" -eq 0 ] && [ ! -s "$STATUS_FILE" ]; then
-    {
-        echo "The analysis API did not answer within 90 seconds of starting."
-        echo "It may still be loading the sentiment models on first boot -- reload shortly."
-        echo
-        echo "--- last lines of the API log ---"
-        tail -n 25 "$API_LOG"
-    } > "$STATUS_FILE"
-    echo "[run.sh] API health check timed out; starting the UI anyway."
-fi
-
-echo "[run.sh] starting Streamlit on :8501 ..."
-exec streamlit run streamlit_app.py --server.port 8501 --server.address 0.0.0.0
+#!/bin/bash
+# Space entrypoint: FastAPI on :7860 (internal) + Streamlit on :8501 (the exposed port).
+#
+# The old version was two lines -- background uvicorn with `&`, then start Streamlit -- and
+# never checked whether the API came up. That matters because config.py raises ImportError
+# when FINNHUB_KEY or REDDIT_CLIENT_ID is unset, which kills uvicorn instantly. Streamlit
+# then starts fine, the Space reports "Running" in green, and every click returns
+# "Failed to connect to the API" with no indication that the actual cause is a missing
+# secret. Same failure shape as every other bug in this codebase: a real error absorbed by
+# the layer above and re-presented as something vague.
+#
+# So: wait for the API to actually answer, and if it does not, write the reason somewhere
+# the UI can read and display. Streamlit still starts either way -- a Space that boots and
+# explains itself is far more useful than one that exits.
+set -u
+
+STATUS_FILE=/tmp/api_status
+API_LOG=/tmp/api.log
+: > "$STATUS_FILE"
+
+echo "[run.sh] starting FastAPI on :7860 ..."
+uvicorn trade_analysis.enhanced_api:app --host 0.0.0.0 --port 7860 > "$API_LOG" 2>&1 &
+API_PID=$!
+
+# python, not curl -- python:3.12-slim ships no curl and adding it just for a health check
+# is a layer nobody needs.
+ready=0
+for _ in $(seq 1 45); do
+    if ! kill -0 "$API_PID" 2>/dev/null; then
+        {
+            echo "The analysis API exited while starting up."
+            echo
+            echo "Most common cause: a required Space secret is not set."
+            echo "Settings -> Variables and secrets: FINNHUB_KEY, REDDIT_CLIENT_ID,"
+            echo "REDDIT_CLIENT_SECRET, REDDIT_USER_AGENT."
+            echo
+            echo "--- last lines of the API log ---"
+            tail -n 25 "$API_LOG"
+        } > "$STATUS_FILE"
+        echo "[run.sh] API process died during startup; see $STATUS_FILE"
+        break
+    fi
+    if python -c "import urllib.request; urllib.request.urlopen('http://localhost:7860/', timeout=2)" 2>/dev/null; then
+        echo "ready" > "$STATUS_FILE"
+        ready=1
+        echo "[run.sh] API is answering."
+        break
+    fi
+    sleep 2
+done
+
+if [ "$ready" -eq 0 ] && [ ! -s "$STATUS_FILE" ]; then
+    {
+        echo "The analysis API did not answer within 90 seconds of starting."
+        echo "It may still be loading the sentiment models on first boot -- reload shortly."
+        echo
+        echo "--- last lines of the API log ---"
+        tail -n 25 "$API_LOG"
+    } > "$STATUS_FILE"
+    echo "[run.sh] API health check timed out; starting the UI anyway."
+fi
+
+echo "[run.sh] starting Streamlit on :8501 ..."
+exec streamlit run streamlit_app.py --server.port 8501 --server.address 0.0.0.0