Upload with huggingface_hub

app.py

@@ -0,0 +1,46 @@
+# # QA
+
+# Questions answering with embeddings.  Adapted from [OpenAI
+# Notebook](https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb).
+
+import datasets
+import numpy as np
+
+from minichain import EmbeddingPrompt, TemplatePrompt, show_log, start_chain
+
+# We use Hugging Face Datasets as the database by assigning
+# a FAISS index.
+
+olympics = datasets.load_from_disk("olympics.data")
+olympics.add_faiss_index("embeddings")
+
+
+# Fast KNN retieval prompt
+
+
+class KNNPrompt(EmbeddingPrompt):
+    def find(self, out, inp):
+        res = olympics.get_nearest_examples("embeddings", np.array(out), 3)
+        return {"question": inp, "docs": res.examples["content"]}
+
+
+# QA prompt to ask question with examples
+
+
+class QAPrompt(TemplatePrompt):
+    template_file = "qa.pmpt.tpl"
+
+
+with start_chain("qa") as backend:
+    question = "Who won the 2020 Summer Olympics men's high jump?"
+    prompt = KNNPrompt(backend.OpenAIEmbed()).chain(QAPrompt(backend.OpenAI()))
+    result = prompt(question)
+    print(result)
+
+# + tags=["hide_inp"]
+QAPrompt().show(
+    {"question": "Who won the race?", "docs": ["doc1", "doc2", "doc3"]}, "Joe Bob"
+)
+# -
+
+show_log("qa.log")

qa.pmpt.tpl

@@ -0,0 +1,12 @@
+Answer the question as truthfully as possible using the provided context, and if the answer is not contained within the text below, say "I don't know."
+
+Context:
+
+{% for doc in docs %}
+* {{doc}}
+{% endfor %}
+
+ Q: {{question}}
+
+ A: 
+

requirements.txt

@@ -0,0 +1,3 @@
+gradio
+git+https://github.com/srush/minichain
+manifest-ml