Consolidate prompt quality roadmap into implementation sequence

This commit is contained in:
2026-11-11 16:46:55 +00:00
parent d15d80c79f
commit 7c567d8d8b

View File

@@ -2,218 +2,203 @@
**Updated:** 2026-11-11
**Status:** Topic 1 resolved in ADR 0003; topics 2–4 and 6 decided, with
implementation pending; topic 5 deferred. No open decisions remain in the
current scope. Topic 5 retains recommendations for future discussion.
**Status:** Decisions settled; implementation pending. Editorial evaluation is
deferred. No open decisions remain in the current scope.
## Baseline and Intended Direction
## Target State and Constraints
The review of revision `6d34602` covered all eleven
[prompt definitions](../internal/prompts.md), their shared instructions, and
the context and response handling for proposals, validation, and repair.
It was a design review, not a measurement of live-model quality.
Audita will use clearer proposal, validation, and repair prompts, actionable
rejection feedback, and no model-reported confidence scores or thresholds.
Repair will reconsider corrections on their merits and revise or withdraw them
based on the supplied evidence. The initial review covered the eleven
[prompt definitions](../internal/prompts.md) and their consumers at revision
`6d34602`; it was a design review, not a live-model quality measurement.
The maintainer reports that nearly all module behavior was selected through
practical experimentation. Treat that experience as evidence when reviewing the
design: overlapping responsibilities and repeated passes are not inherently
redundant. The deferred evaluation work does not imply an absence of empirical
support for the existing module design. This experience informs ongoing choices;
it does not lock in the current default. Frequently reevaluate the default module
combination and sequence, selecting what has been observed to work best for most
workloads.
Preserve the following throughout implementation:
Retain the strengths: conservative transcript editing, explicit literal-span and
deletion semantics, distinct glossary context and protection, full before/after
validation, independent repair conversations with withdrawal, and native
PromptKit/ConfigKit prompt replacement. Improve consistency across the three
roles, useful rejection feedback, and maintainability. Remove model-reported
confidence scores and confidence thresholds throughout the correction workflow.
- **Editorial acceptance:** module focus guides requested work, while validators
permit otherwise acceptable minor editorial cleanup across modules. Follow
[ADR 0003](../adr/0003-module-focus-and-editorial-acceptance.md), which records
the maintainer's finding that strict scope instructions also suppress desirable
edits. Applicability, protection, and other substantive checks still apply.
- **Correction mechanics:** retain literal-span and deletion semantics, distinct
glossary context and protection, full before/after validation, and independent
repair conversations with full history, withdrawal, and revalidation. Preserve
native PromptKit/ConfigKit prompt replacement and the existing module structure.
- **Module behavior:** keep `grammar` and its current requested scope; retain
glossary and homophone modules with their useful overlap. Keep the current
[default sequence](../config.md#pipeline-and-glossary-selections) for this pass
and preserve configurable additional passes.
- **Language:** retain English as the bundled default and useful existing campaign
examples, supplemented with a few general examples. Broader language support is
outside this pass.
- **Meaning reversal:** preserve the narrowly tuned guard against reversal of the
full segment's assertion. Its targeted prompt changes are specified below.
## Discussion and Work Order
Nearly all module behavior reflects the maintainer's practical experimentation.
Two glossary passes have proved useful, and some workloads benefit from three or
more. These observations inform the current choice without fixing a permanent
number or order of passes. Frequently reevaluate the default module combination
and sequence and select what has been observed to work best for most workloads.
This ongoing maintenance does not depend on the deferred formal evaluation work.
### 1. Module Focus and Editorial Acceptance — Resolved
## Implementation Sequence
The intentional distinction between requested module focus and acceptance of
minor editorial cleanup is recorded in
[ADR 0003](../adr/0003-module-focus-and-editorial-acceptance.md). The maintainer's
testing found that strict scope instructions also suppress desirable edits.
Apply the accepted distinction throughout the remaining prompt work.
Work through the following steps in order. Keep schemas, consumers, prompt
examples, and affected tests consistent as each change lands, and update the
canonical documentation with the corresponding behavior. The final step checks
the assembled result. Proceed without waiting for live-model evaluation evidence
or a captured baseline.
### 2. Make Rejection Reasons Useful for Repair — Decided
### 1. Remove Confidence Throughout the Workflow
Validator explanations now become repair feedback, but current instructions give
little guidance about producing actionable reasons.
Remove confidence entirely from proposals, validation, and repair. Audita has no
established calibration that gives its model-reported scores reliable numerical
meaning across transcripts and models. Proposal scores currently affect
acceptance; validator scores are range-checked but do not determine approval.
Retaining the numbers adds complexity and encourages repair toward a threshold.
**Decision:** retain the existing `reason` field. Instruct validators to explain
rejections with a concise description of the specific problematic edit, the
applicable criterion, and relevant supplied evidence, within their own remit.
For editorial review, for example: removing "might" turns a tentative statement
into a definite one.
Validators identify the defect; the repair module owns revision or withdrawal.
Respect the accepted minor editorial allowance, distinguish insufficient
evidence from an established defect, and include distinct material issues when
found without requiring an exhaustive analysis.
Approvals require no explanation: instruct the model to return `reason: ""`.
The existing schema requires the field but permits an empty string, which the
response handler already accepts. This reason convention itself needs no schema
or response-handling changes; topic 3 separately removes confidence from those
contracts. The rejection guidance is a prompt instruction;
this decision adds no runtime rule rejecting blank explanations.
**Remaining work:** update the shared validation guidance and all three validator
definitions with representative rejection examples and the empty-reason approval
convention. Check rendered prompts and representative repair payloads to confirm
that rejection feedback identifies defects without prescribing replacement
corrections. Consider schema expansion only if a concrete need emerges.
### 3. Reconsider Corrections on Their Merits and Remove Confidence — Decided
**Decision:** remove confidence entirely from proposals, validation, and repair,
including the threshold validator and its configuration. Audita has no established
calibration that gives its model-reported scores a reliable numerical meaning
across transcripts and models. Proposal scores currently affect acceptance;
validator scores are range-checked but do not determine approval. Retaining these
numbers adds complexity and encourages repairs aimed at meeting a threshold.
The target workflow relies on substantive validation criteria and supplied
evidence, without a replacement numerical or categorical confidence scale.
Repair instructions should ask the model to reassess a correction using the
original transcript, supplied context, and rejection feedback. Revise when the
evidence supports a correction that addresses the identified problems; withdraw
when the proposal remains unsupported, unnecessary, or too uncertain. Withdrawal
is a successful resolution that preserves the original text. Explain concrete
uncertainty in rejection feedback where relevant, such as context that cannot
distinguish two possible names. Preserve the module's requested focus, the
accepted minor editorial allowance, the full-history conversation model, and
the existing repair lifecycle and revalidation.
**Remaining work:**
- Remove confidence fields, instructions, and examples from all proposal,
validation, and repair prompts and response schemas, and from their Go models,
decoding, and validation paths.
- Remove confidence fields from response schemas, Go models, decoding, and
validation paths, together with corresponding prompt instructions and examples.
- Remove the confidence-threshold validator, its registration and chain entries,
root and stage threshold settings, and any corresponding CLI exposure.
- Remove confidence-specific rejection feedback, reason codes, report/debug
fields, and other plumbing wherever present. Update affected tests, fixtures,
configuration examples, and canonical documentation when implementation lands.
- Treat this as a deliberate contract change for configuration, custom prompts,
and affected serialized outputs; provide migration guidance to remove obsolete
fields and settings. Do not retain inert confidence fields or threshold knobs.
- Verify that confidence-free proposals, validation decisions, and repair
responses work end to end, while substantive validators, revision, withdrawal,
and revalidation continue to operate as intended.
fields, and other supporting code wherever present. Update affected tests,
fixtures, and configuration examples.
- Treat removal as a deliberate contract change for configuration, custom prompts,
and affected serialized outputs. Retain no inert confidence fields or threshold
knobs, and introduce no replacement numerical or categorical confidence scale.
Removing the proposal threshold can admit corrections it previously rejected.
The deferred evaluation work in topic 5 can assess that effect when resumed;
removal is the selected direction and does not depend on that work.
That effect can be assessed when editorial evaluation resumes; it does not block
the selected removal.
### 4. Simplify Instructions and Improve Examples — Decided
### 2. Improve Validation Feedback
Restrictions and examples recur across proposal, validation, and repair assets;
the grammar prompt repeats its word-substitution boundary several ways.
Retain the existing `reason` field. Update shared validation guidance and all
three validator definitions to request concise rejection explanations identifying
the specific problematic edit, applicable criterion, and relevant supplied
evidence within each validator's remit. For editorial review, for example:
removing "might" turns a tentative statement into a definite one.
Validators identify defects; the repair module owns revision or withdrawal.
Distinguish insufficient evidence from an established defect, respect the accepted
editorial allowance, and include distinct material issues when found without
requiring exhaustive analysis. Provide representative rejection examples.
Approvals need no explanation: instruct the model to return `reason: ""` and use
that convention in approval examples. The field remains required and permits an
empty string; this reason convention itself needs no schema or response-handling
change. Explanation guidance is expressed in prompts, with no new runtime
rule rejecting blank explanations. Consider schema expansion only if a concrete
need emerges.
The meaning-reversal validator requires particular care. The maintainer has
observed overly broad instructions causing it to reject virtually all corrections.
Preserve its core instructions and make only small, targeted feedback and example
changes, alongside confidence removal and the approval-reason convention:
- Detect reversal of the full segment's assertion, not every change in meaning.
Changes to names, referents, or other information alone are not grounds for
rejection. "We met Alice" becoming "We met Alex" illustrates a meaning change
that is not a reversal.
- Assess the full before/after segment. An antonym appearing in the replacement
does not by itself establish reversal.
- Instruct rejection feedback to identify the specific opposition in the
assertions; "this changes the meaning" is insufficient.
- Approve corrections outside this narrow rejection criterion, leaving other
validators to assess their own criteria. Preserve the guard even though it can
exclude a genuine mistranscription correction that reverses the assertion;
no contextual exception is selected in this pass.
### 3. Refine Repair Guidance
Ask the model to reassess the correction using the original transcript, supplied
context, and rejection feedback:
- Revise when the evidence supports a correction that addresses the identified
problems.
- Withdraw when the proposal remains unsupported, unnecessary, or too uncertain.
Withdrawal is a successful resolution that preserves the original text.
Use concrete explanations of uncertainty where relevant, such as context that
cannot distinguish two possible names. Preserve requested module focus and the
accepted editorial allowance. Apply these instructions within the existing
full-history repair lifecycle and revalidation, without adding new retry rules.
### 4. Consolidate Prompts, Descriptions, and Examples
Make a focused editing pass using the existing prompt fragment and module
structure. Organize instructions around shared contracts (including response
format, literal matching, and deletion semantics), module-specific requested work,
and the distinct responsibilities of validation and repair.
**Decision:** make a focused editing pass using the existing module and prompt
fragment structure. Organize instructions around shared contracts (including
response format, literal matching, and deletion semantics), module-specific
requested work, and the distinct responsibilities of validation and repair.
Consolidate duplicated rules where small shared fragments prevent drift, remove
contradictions and unnecessary repetition, and keep each rendered prompt
understandable on its own. Retain repetition across separate model calls when
each call needs the instruction.
each call needs the instruction. Preserve the meaning-reversal validator's tuned
core wording as specified in step 2; broad simplification for consistency is not
appropriate there. Shorter prompts alone are not the objective.
Apply the meaning-reversal decision in topic 6 as an exception to broad prompt
simplification: preserve that validator's tuned core instructions and make only
small, targeted changes to feedback and examples, alongside the agreed confidence
removal and approval-reason convention.
Clarify descriptions and examples as follows:
Use short contextual contrasts to clarify difficult judgments, such as an
accidental restart versus deliberate emphasis, or a glossary match supported by
context versus a merely similar spelling. Examples should illustrate how context
informs the decision, not create unconditional editing rules. Preserve the
distinction between requested module focus and acceptance of otherwise appropriate
minor editorial cleanup. Shorter prompts alone are not the objective.
- Keep the `grammar` name, module key, and prompt IDs. Describe its requested work
as conservative punctuation, capitalization, spacing, and whole-word `a`/`an`
article cleanup. Preserve the current scope and cross-module editorial allowance.
- Describe glossary stages as focusing on known terms supported by selected
glossaries, and homophone stages as looking more broadly for plausible
spoken-word mistranscriptions, including glossary-supported corrections. Do not
exclude an otherwise supported homophone correction because it targets a
glossary term.
- Use short contextual contrasts for difficult judgments, such as accidental
restarts versus deliberate emphasis and context-supported glossary matches
versus merely similar spellings. Examples illustrate judgments rather than
establish unconditional editing rules.
- Retain useful campaign examples and add a few general examples. Use broadly
applicable wording for names, terminology, and glossary entries without asserting
D&D specificity or adding a disclaimer about it.
**Remaining work:** revise the proposal, validation, and repair assets and shared
fragments consistently with topics 1–3, retaining the existing module structure.
Check rendering, response-schema compatibility, and integration behavior as
appropriate. Proceed with these revisions before waiting for live-model
evaluation evidence or a captured baseline. Topic 5 is deferred and is not a
prerequisite for this editing pass.
### 5. Complete Integration Checks and Documentation
### 5. Evaluate Editorial Quality — Deferred
Follow the [testing policy](../policy/testing.md) and the
[development validation guidance](../development.md#baseline-validation). Verify
the assembled workflow, including:
**Decision:** defer this work for now. Evaluation corpus preparation, live-model
comparisons, and an evaluation workflow are outside the current implementation
scope. This deferral does not block topics 2–4 or the remaining product
discussions; ordinary implementation checks still apply. The following direction
is retained for reconsideration when evaluation work resumes.
- Confidence-free proposal, validation, and repair responses work through their
schemas and consumers; substantive validators remain effective in the workflow.
- Rendered prompts and examples match the revised contracts, including empty
approval reasons and rejection feedback carried into repair requests.
- Revision, withdrawal, and full revalidation retain their existing mechanics.
Rendering, schema, and controlled-response tests establish integration behavior,
but do not establish the quality of model editorial judgments.
Review the actual prompt wording and representative repair payloads to check the
selected responsibilities and the narrow meaning-reversal criterion. Offline
tests verify contracts and integration; they do not establish live-model
editorial quality. Complete the appropriate package checks, build, and full test
suite for the changes, without introducing paid calls into the default suite.
**Future recommendation:** assemble a modest evaluation set containing both needed edits
and text that should remain unchanged. Include ambiguous glossary matches,
intentional repetition, meaningful uses of apparent filler, punctuation that
changes meaning, and repairs that should be withdrawn. Compare unnecessary edits,
missed corrections, validator disagreement, and repair outcomes using independent
editorial assessment; validator approval alone is insufficient evidence.
Include a comparison against the prior confidence-threshold behavior to identify
changes in both useful corrections and erroneous edits after confidence removal.
Update canonical [configuration](../config.md), [CLI](../cli.md),
[prompt](../internal/prompts.md), [module](../internal/modules.md),
[validator](../internal/validators.md), and
[output](../integrations/output-files.md) documentation where affected, together
with maintained examples. Provide migration guidance to remove obsolete confidence
settings and fields from configuration, custom prompts, and affected output
consumers. Check links, examples, and whitespace, and remove stale descriptions
of confidence-based behavior.
When this work resumes, discuss representative models, expected outcomes, and
how to run evaluations. Use the prior repository revisions for
before/after comparisons when that work proceeds; a pre-change evaluation run is
not required. Keep model-dependent evaluation separate from the default offline
test suite; no numerical quality gate is selected here.
## Deferred Editorial Evaluation
### 6. Revisit Broader Product Choices
Evaluation corpus preparation, live-model comparisons, and an evaluation workflow
are outside this implementation scope. No baseline run or numerical quality gate
is required before the agreed changes proceed.
The following decisions guide this pass without waiting for the deferred
evaluation work:
When this work resumes, consider a modest collection of needed corrections and
text that should remain unchanged: ambiguous glossary matches, intentional
repetition, meaningful apparent filler, punctuation affecting meaning, and repairs
that should be withdrawn. Compare unnecessary edits, missed corrections,
validator disagreement, and repair outcomes using independent editorial judgment;
validator approval alone is insufficient evidence.
1. **Grammar naming and scope — Decided:** retain the `grammar` name, module key,
and corresponding prompt IDs. Clarify its description as conservative
punctuation, capitalization, spacing, and whole-word `a`/`an` article cleanup.
Preserve its current requested scope and the accepted cross-module minor
editorial allowance. Update the relevant module descriptions and canonical
documentation during implementation; no rename or broader grammar-correction
remit is planned.
2. **Glossary/homophone overlap and defaults — Decided:** retain both modules and
their useful overlap. Keep the existing default sequence documented in
[configuration](../config.md#pipeline-and-glossary-selections) for this pass.
Frequently reevaluate the default and set it to the combination and sequence
of modules observed to work best for most workloads. The maintainer has
observed that two glossary passes help and some workloads benefit from three
or more; these observations inform the current choice without fixing a
permanent number or order of passes. Preserve operators' ability to configure
additional passes. Clarify the complementary requested roles: glossary stages
focus on known terms supported by selected glossaries; homophone stages look
more broadly for plausible spoken-word mistranscriptions and may also use
glossary evidence. Do not exclude an otherwise supported homophone correction
merely because it targets a glossary term. No module consolidation is planned
in this pass. Reevaluating defaults does not depend on the deferred formal
evaluation work in topic 5.
3. **Language and examples — Decided:** retain English as the bundled default.
Keep useful existing campaign examples and add a few more general examples
during the prompt editing pass. Use broadly applicable wording for names,
terminology, and glossary entries, without asserting D&D specificity or
adding a disclaimer about it. Broader language support is outside this pass.
4. **Meaning-reversal policy — Decided:** retain the narrow guard against reversal
of the full segment's assertion, rather than rejection of any change in
meaning. The maintainer has observed that overly broad instructions cause this
validator to reject virtually all corrections; its wording requires unusually
precise tuning. Preserve the existing core instructions and make small,
targeted changes, principally improving rejection feedback and adding a
contrasting example. Changes to names, referents, or other information are
not grounds for rejection by this validator unless they reverse the assertion:
"We met Alice" becoming "We met Alex" changes meaning but is not a reversal.
Assess the full before/after segment; the presence of an antonym in a
replacement does not by itself establish reversal. Instruct rejection feedback
to identify the specific opposition in the assertions; "this changes the
meaning" is insufficient. Approve corrections outside this narrow rejection
criterion, leaving other validators to assess their own criteria. These are
prompt instructions, not new runtime checks on explanation text. Preserve the
safeguard even though it can exclude a genuine mistranscription correction
that reverses the assertion; no contextual exception is selected in this pass.
Discuss representative models, expected outcomes, and an evaluation workflow at
that time. Prior repository revisions can support before/after comparisons,
including the effects of removing confidence thresholds. Keep model-dependent
evaluation separate from the default offline test suite.