llmEngine draft version (proof of concept)

[prabinrs]

Added LLM engine

• get notes from observation_facts()
• call external llm openAI API compliant and check for the presence of given term
• add the terms as observatuion_facts()

Files changes :

• llmEngine
• llmHelper
• llmAPI
• i2b2_cdi_app
• test files

• test_llm_api
• Test_llm_engine

tested on the mock dataset

SET search_path = i2b2demodata;

-- 1. Add the patient
INSERT INTO patient_dimension (patient_num, sex_cd, age_in_years_num, sourcesystem_cd)
VALUES (99001, 'M', 45, 'TEST_LLM')
ON CONFLICT DO NOTHING;

-- 2. Add a patient_mapping entry (required for the fact loader)
INSERT INTO patient_mapping (patient_ide, patient_ide_source, patient_num, patient_ide_status, project_id, sourcesystem_cd)
VALUES ('99001', 'TEST_LLM', 99001, 'A', 'demo', 'TEST_LLM')
ON CONFLICT DO NOTHING;

-- 3. Add the note concept
INSERT INTO concept_dimension (concept_path, concept_cd, name_char, sourcesystem_cd)
VALUES ('/i2b2/Notes/ClinicalNotes/', 'NOTE:CLINICAL', 'Clinical Notes', 'TEST_LLM')
ON CONFLICT DO NOTHING;

-- 4. Add the output concept (parent only; child codes are auto-created by llmEngine)
INSERT INTO concept_dimension (concept_path, concept_cd, name_char, sourcesystem_cd)
VALUES ('/i2b2/LLM/NLP/Extract/', 'LLM:NLP:FOUND', 'LLM NLP Extraction', 'TEST_LLM')
ON CONFLICT DO NOTHING;

-- 5. Insert the clinical note as a fact
INSERT INTO observation_fact
    (encounter_num, patient_num, concept_cd, provider_id, start_date,
     modifier_cd, instance_num, observation_blob, sourcesystem_cd)
VALUES
    (1, 99001, 'NOTE:CLINICAL', '@', '2024-01-01 00:00:00',
     '@', 1,
     'Patient is a 45-year-old male with type 2 diabetes. Currently on metformin 500mg twice daily. Last HbA1c was 7.2%.',
     'TEST_LLM')
ON CONFLICT DO NOTHING;