feat: baseline RAG pipeline with FunSQL grounding, ingestion, and HuggingFace model support

.CondaPkg/.gitattributes

@@ -0,0 +1,4 @@
+# SCM syntax highlighting & preventing 3-way merges
+pixi.lock merge=binary linguist-language=YAML linguist-generated=true -diff
+# SCM syntax highlighting & preventing 3-way merges
+pixi.lock merge=binary linguist-language=YAML linguist-generated=true -diff

.CondaPkg/.gitignore

@@ -0,0 +1,6 @@
+# pixi environments
+.pixi/*
+!.pixi/config.toml
+# pixi environments
+.pixi/*
+!.pixi/config.toml

.CondaPkg/.pixi/config.toml

@@ -0,0 +1 @@
+detached-environments = false

.CondaPkg/pixi.toml

@@ -0,0 +1,14 @@
+[dependencies]
+openssl = ">=3, <3.6"
+
+    [dependencies.python]
+    channel = "conda-forge"
+    build = "*cp*"
+    version = ">=3.10,!=3.14.0,!=3.14.1,<4"
+
+[workspace]
+name = ".CondaPkg"
+platforms = ["win-64"]
+channels = ["conda-forge"]
+channel-priority = "strict"
+description = "automatically generated by CondaPkg.jl"

Project.toml

@@ -4,25 +4,37 @@ version = "0.1.0"
 authors = ["ParamThakkar123 <paramthakkar864@gmail.com> and TheCedarPrince <jacobszelko@gmail.com>"]
 
 [deps]
+CondaPkg = "992eb4ea-22a4-4c89-a5bb-47a3300528ab"
+DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+DrWatson = "634d3b9d-ee7a-5ddf-bec9-22491ea816e1"
+DuckDB = "d2f5444f-75bc-4fdf-ac35-56f514c445e1"
+FunSQL = "cf6cc811-59f4-4a10-b258-a8547a8f6407"
 HuggingFaceHub = "d0076355-e2c0-48e6-a044-05906e51b7fc"
 JSON3 = "0f8b85d8-7281-11e9-16c2-39a750bddbf1"
 LibPQ = "194296ae-ab2e-5f79-8cd4-7183a0a5a0d1"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 PromptingTools = "670122d1-24a8-4d70-bfce-740807c42192"
+PythonCall = "6099a3de-0909-46bc-b1f4-468b9a2dfc0d"
 RAGTools = "16ddad29-bbe8-45a7-857d-3d9514eb0023"
 Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
 [compat]
 Aqua = "0.4"
+CondaPkg = "0.2.36"
+DataFrames = "1.8.0"
+DrWatson = "2.19.1"
+DuckDB = "1.3.2"
+FunSQL = "0.15.0"
 HTTP = "1.11.0"
 HuggingFaceHub = "0.1.2"
 JSON3 = "1.14.3"
 LibPQ = "1.18.0"
 LinearAlgebra = "1.10"
-PromptingTools = "0.82.1"
-RAGTools = "0.7.0"
+PromptingTools = "≥ 0.82.1, < 0.93"
+PythonCall = "0.9.35"
+RAGTools = "≥ 0.7.0, < 0.11"
 Serialization = "1.10"
 SparseArrays = "1.10"
 Statistics = "1.10"

demo/basic_rag_pipeline.jl

@@ -0,0 +1,65 @@
+using DrWatson
+@quickactivate "HealthLLM"
+using HealthLLM
+using PromptingTools
+using RAGTools
+
+println("=" ^ 60)
+println("HealthLLM.jl — Baseline RAG Pipeline Demo (HuggingFace Native)")
+println("=" ^ 60)
+
+model_embedding = "hf:sentence-transformers/all-mpnet-base-v2"
+model_name = "hf:Qwen/Qwen2.5-Coder-1.5B-Instruct"
+
+HF_TOKEN = get(ENV, "HF_API_TOKEN", get(ENV, "HUGGING_FACE_HUB_TOKEN", ""))
+if isempty(HF_TOKEN)
+    @warn "HF_API_TOKEN not set. Set it to use HuggingFace Inference API."
+    println("  See: https://huggingface.co/settings/tokens")
+end
+HealthLLM.configure_hf_token!(; token=isempty(HF_TOKEN) ? nothing : HF_TOKEN)
+
+println("\n[1/4] Registering models and schemas...")
+emb_schema = HealthLLM.Utils.get_schema(nothing, model_embedding)
+gen_schema = HealthLLM.Utils.get_schema(nothing, model_name)
+println("  Embedding schema: ", typeof(emb_schema))
+println("  Generator schema: ", typeof(gen_schema))
+
+PromptingTools.register_model!(name=model_name, schema=gen_schema)
+PromptingTools.register_model!(name=model_embedding, schema=emb_schema)
+PromptingTools.MODEL_CHAT = model_name
+PromptingTools.MODEL_EMBEDDING = model_embedding
+
+println("\n[2/4] Building grounding index from curated docs...")
+index = HealthLLM.build_grounding_index(; embedder_model=model_embedding, verbose=true)
+n_chunks = length(RAGTools.chunks(index))
+println("  $n_chunks chunks indexed from grounding corpus.")
+
+println("\n[3/4] Registering FunSQL prompt template...")
+HealthLLM.register_funsql_template!()
+println("  Template :FunSQLQueryGeneration registered.")
+
+questions = [
+    "Count patients per care site place of service",
+    "List patient counts by age and gender for patients with hip fracture",
+]
+
+println("\n[4/4] Running RAG queries...")
+for (i, q) in enumerate(questions)
+    println("\n" * "-" * 60)
+    println("Query $i: $q")
+    println("-" * 60)
+    try
+        result = HealthLLM.answer_question(q, index, model_embedding, model_name; verbose=true)
+        answer_text = result isa RAGTools.RAGResult ? result.final_answer : string(result)
+        println("\nFinal answer:\n$answer_text")
+    catch e
+        println("Error during query: $e")
+        if e isa MethodError
+            showerror(stdout, e, catch_backtrace())
+        end
+    end
+end
+
+println("\n" * "=" * 60)
+println("Demo complete.")
+println("=" * 60)