haipipe-end-input2src

name	haipipe-end-input2src
description	Input2SrcFn specialist -- designs/reviews the wire-payload->record function in an Endpoint_Set (deserializes a JSON request into a ProcessedDF row). Platform-specific: one impl per deploy platform (SageMaker flat JSON vs Databricks dataframe_records); --platform picks (default sagemaker). Called by /haipipe-end when intent references Input2SrcFn, payload-to-record deserialization, or `input2src`.
argument-hint	[verb] [use_case] [--platform sagemaker|databricks] [args...]
allowed-tools	Bash, Read, Write, Edit, Grep, Glob
metadata	{"version":"0.2.1","last_updated":"2026-07-08","summary":"Input2SrcFn specialist — design and review of the wire-payload-to-record function in an Endpoint_Set (deserializes JSON request into a ProcessedDF row)."}

Skill: haipipe-end-input2src

Per-Fn-type specialist for Input2SrcFn — the inference function that deserializes a wire-format JSON payload back into a ProcessedDF row. The "inbound" half of the wire I/O pair (paired with Src2InputFn). One of the 5 inference Fn-types inside an Endpoint_Set. See ref/concepts.md for Input2SrcFn semantics.

Verb axis: design | review | list | concepts Use case: each Input2SrcFn impl is scoped to ONE decoding (CGM, weight, generic inference). design and review take a use_case argument. Platform: ONE Input2SrcFn PER PLATFORM per use-case (owner decision 2026-07-05, supersedes LESSON L16): a SageMaker payload gets a SageMaker impl (flat JSON); a Databricks payload gets its own impl that owns the {'dataframe_records': [...]} envelope. Put the platform in the impl name (e.g. Databricks; sagemaker unmarked). Keep variants thin over a shared body; the duplicate-fix cost is accepted (L16 history + supersession note). --platform sagemaker|databricks picks which platform's Fn a design/review targets (default sagemaker).

---

Commands

/haipipe-end-input2src                                                  -> show Input2SrcFn ref
/haipipe-end-input2src concepts                                         -> same
/haipipe-end-input2src list                                             -> list use-case impls
/haipipe-end-input2src design <use_case> [endpoint_set] [--platform sagemaker|databricks]  -> design that platform's Fn
/haipipe-end-input2src review <use_case> [endpoint_set] [--platform sagemaker|databricks]  -> structural audit of that platform's Fn
# --platform selects WHICH platform's impl to design/review (default sagemaker);
# one Fn per platform per use-case (owner decision 2026-07-05)

Use cases (concrete impls in code/haifn/fn_endpoint/fn_input2src/, as of 2026-04-25)

InferenceV240727                                generic inference         (sagemaker)
CGMExamples_v260101                             CGM examples decoder      (sagemaker)
WellDocWeight_Payload2Src_v260305               weight decoder            (sagemaker)
WellDocWeight_OldFormat2Src_v260318             weight decoder (legacy)   (sagemaker)
<YourDatabricksFn>                                🚩  generic Databricks         (databricks)
CGMDecoder_Databricks_Payload2Src_v260101          🚩  CGM Databricks variant     (databricks)

🚩 = Databricks-platform impl (one wire-Fn per platform by design; sagemaker unmarked)

If <use_case> is omitted, the skill should Bash("ls code/haifn/fn_endpoint/fn_input2src/") and confirm the impl with the user before proceeding (the snapshot above goes stale; disk is the truth).

---

Dispatch Table

Verb       Reads
---------- ------------------------------------------------------------------
design     ref/concepts.md
           ../haipipe-end/fn/fn-design.md
           ../haipipe-end/ref/0-overview.md
review     ref/concepts.md
           ../haipipe-end-endpointset/fn/fn-review.md
concepts   ref/concepts.md  (only)

---

Step-by-Step Protocol

Step 0: Read ref/concepts.md — Input2SrcFn semantics, schema validation, decoding rules. Step 1: For design, also read ../haipipe-end/fn/fn-design.md + ../haipipe-end/ref/0-overview.md. For review, also read ../haipipe-end-endpointset/fn/fn-review.md. Step 2: Execute the procedure scoped to Input2SrcFn. Must stay in sync with its inverse pair /haipipe-end-src2input. Step 3: Emit the structured tail.

---

Scope

Owns:

• Input2SrcFn concept ref (ref/concepts.md)
• Input2SrcFn design + review scoped to ONE Fn-type

Does NOT own:

• Inverse Fn — see /haipipe-end-src2input
• Other 3 Fn-types — -meta, -trig, -post
• Whole-artifact verbs — /haipipe-end-endpointset
• Deployment — /haipipe-end-deploy-*

Pair invariant

For any record R and platform P: Input2SrcFn_P(Src2InputFn_P(R)) == R (the SAME-platform pair must roundtrip). Changes here typically require a paired update in -src2input for the same platform.

Roundtrip test (REQUIRED for design and review)

Every design or review MUST include a roundtrip test against real example data from the ModelInstanceStore. See the paired skill haipipe-end-src2input for the full test protocol.

The test verifies:

1. All non-empty source tables survive the roundtrip
2. Features produced from original vs reconstructed data are identical
3. Model predictions match within tolerance (< 0.001)

Why this matters: Input2SrcFn must reconstruct ALL tables that CaseFns read — not just the ones that seem important. If Src2InputFn serializes only 4 of 19 tables and Input2SrcFn creates empty stubs for the rest, CaseFns that read the missing tables produce zero features → different predictions. This is silent — no error, just wrong scores.

The builder script (e1_build_* in the endpoint fn_develop task folder) must include this test paired with the Src2InputFn. If the roundtrip fails, neither Fn is production-ready.