[jonahgabriel]

Summary

Ticket OMN-20372, lane step 1 of 3: fill the omnibase_infra price list
src/omnibase_infra/configs/pricing_manifest.yaml, the single pricing authority per the
operator-approved direction. Steps 2 (omniclaude cost_accounting.py) and 3 (omnimarket
cost_pricing and pricing.py) are separate PRs and wait for a released omnibase_infra
carrying these entries; the newest tag, v0.38.61, does not carry them.

Changes

• Added paid entries (USD per 1k tokens, input/output), with source and date:
  • claude-haiku-4-5: 0.001 / 0.005
  • claude-sonnet-4-6: 0.003 / 0.015
  • claude-opus-4-5: 0.005 / 0.025
  • Source: https://platform.claude.com/docs/en/about-claude/pricing, read 2026-10-02
  • gemini-2.5-flash: 0.0003 / 0.0025
  • gemini-2.5-flash-lite: 0.0001 / 0.0004
  • Source: https://ai.google.dev/gemini-api/docs/pricing, Standard paid tier text input, read 2026-10-02
  • Each entry carries evidence.source_url and retrieved_at, and is marked
    LOW_CONFIDENCE / FALLBACK_PROVIDER_DOCUMENTATION.
• Added zero-cost local entries: qwen3.8, Qwen3.8-27B, Qwen3-Coder-30B,
  DeepSeek-R1-14B — the local ids omnimarket cost_pricing.yaml prices.
• No existing entry was changed.
• New test: tests/unit/configs/test_pricing_manifest_fill_omn20372.py.

Verification

Run locally:

• scripts/lint_pricing_manifest.py passes.
• pytest on tests/ci/test_pricing_manifest_schema.py, tests/unit/configs,
  tests/unit/models/pricing, and tests/unit/nodes/node_savings_estimation_compute:
  157 passed.
• ruff check and ruff format clean.
• pre-commit passed on commit.

Follow-ups

Found during this work, not changed here:

• (a) Existing claude-opus-4-6 is 0.015 / 0.075 per 1k, but the Anthropic pricing page
  lists Opus 4.6 at $5 input / $25 output per MTok (0.005 / 0.025 per 1k); omniclaude
  cost_accounting.py carries the same 15/75.
• (b) omniclaude prices claude-haiku-4-5 at $0.80 / $4.00 per MTok; the page says
  $1 / $5.
• (c) omnimarket cost_pricing.yaml prices gemini-2.5-flash at 0.10 / 0.40 per MTok and
  gemini-2.5-flash-lite at 0.075 / 0.30; the Google page says 0.30 / 2.50 and 0.10 / 0.40.
  Pointing those tables at the manifest in step 3 changes those figures.
• (d) Manifest key qwen3.6-35b-a3b differs in case from omnimarket Qwen3.6-35B-A3B;
  get_entry is exact-match.

Not in scope: the CI guard against new price literals (OMN-19980, Lakshman).

Evidence-Ticket: OMN-20372
Evidence-Source: OCC#12394

Lab pool readback (OMN-18893)

LAB PROOF PASS: omnibase_infra#4480 head 86ca01b559 on lab-101 (192.168.86.101), isolated project omnibase-infra-local, 2026-10-02T19:59:31Z to 2026-10-02T20:11:26Z
  ok   head_matches
  ok   stack_built
  ok   image_identity
  ok   health_8085
  ok   health_8086
  ok   migration_gate_healthy
  ok   no_wiring_failures
  ok   focused_tests
  restored=yes (containers=0 volumes=0 networks=0 images=0 listeners=0 workdir=gone, snapshot-diff=0)
  method: scripts/runtime_build/prepr_runtime_pool.py (OMN-18893)
  receipt OmniNode-ai/omnibase_infra#4480@86ca01b559d7fc0075fbc1a6a926a2fc3bb8cac3:omnibase_infra.runtime_and_scripts@1 result=PASS verifier=MISSING_MANDATORY_CHECK (mandatory passing check(s) missing: ['changed_path_live', 'forwarder_refused_topic']) file=/Users/jonah/.local/state/omni/landing/lab/workers/000059/lab-proof-receipts/omnibase_infra-4480-86ca01b559d7fc0075fbc1a6a926a2fc3bb8cac3.receipt.json event=/Users/jonah/.local/state/omni/landing/lab/workers/000059/lab-proof-receipts/omnibase_infra-4480-86ca01b559d7fc0075fbc1a6a926a2fc3bb8cac3.event.json
  ledger: RELEASE re=<your HOLD id> surface=101-runtime result=PASS restored=yes

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: qwen3-review, gpt-oss-review
Models failed: none
New finding threads: 0
Deduped (already posted on this PR): 0
Below quorum (one model only, reported not threaded): 8
Nit-level findings suppressed: 0

The model is the FINDER, never the gate: merge is gated only by the
deterministic Hostile Review Thread Gate, which blocks while
hostile-reviewer threads are unresolved. Resolve each thread after
addressing (or rejecting, with a reply) its finding.

Below quorum: 8 finding(s) raised by one model only (OMN-18479)

These are reported and NOT dropped, but they get no thread and do not block: a single model's finding no other model reproduced is not evidence enough to stop a merge. Read them; act on them if they are right.

• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — Test asserts retrieved_at equals effective_date, but they are not semantically equivalent | The test test_paid_models_have_documented_rates asserts that raw_entry["evidence"]["retrieved_at"] == raw_en
• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — No test validates that local model entries have the expected evidence structure | The test test_local_models_have_zero_api_cost checks that input and output costs are zero and that the source is LOCAL
• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — Cost estimate test does not account for potential rounding or truncation in estimate_cost | The test test_haiku_cost_estimate_uses_the_documented_rate asserts that estimate_cost("claude-haiku-4-5", 10
• [MAJOR] src/omnibase_infra/configs/pricing_manifest.yaml (gpt-oss-review) — Missing evidence block for local model entries | The newly added local model entries (e.g., qwen3.8, Qwen3.8-27B, Qwen3-Coder-30B, DeepSeek-R1-14B) omit the required "evidence" section that other part
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — No test verifies that the YAML manifest is valid YAML and parseable by the loader | The tests rely on ModelPricingTable.from_yaml() and yaml.safe_load to parse the manifest. If the YAML has a syntax e
• [MINOR] src/omnibase_infra/configs/pricing_manifest.yaml (qwen3-review) — Model IDs use inconsistent casing and naming conventions | The manifest introduces model IDs with mixed casing: qwen3.8 (lowercase), Qwen3.8-27B (PascalCase), Qwen3-Coder-30B (PascalCase), DeepSeek-R1
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — No test validates that the cost estimate for a local model is zero | The test suite verifies that local model entries have zero costs in the YAML, but does not test that ModelPricingTable.estimate_cos
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (gpt-oss-review) — Assumption about return type of estimate_cost in test | The test test_haiku_cost_estimate_uses_the_documented_rate accesses estimate.estimated_cost_usd. If ModelPricingTable.estimate_cost return

[github-actions]

✅ Hostile Reviewer — REVIEWED

Critical findings: 0
Major findings: 3
Total findings: 11
Models succeeded: qwen3-review,gpt-oss-review
Models unavailable: none

Action required: findings were posted as review threads. Address or reject each one, then resolve its thread — the Hostile Review Thread Gate fails while hostile-reviewer threads are unresolved (OMN-17492).

---

Semantics (OMN-17492 — the model finds, thread resolution gates)

Surface	Meaning	Blocks merge?
Review threads	Per-finding, posted by the reviewer	No (informational)
Hostile Review Thread Gate	Deterministic: unresolved hostile-reviewer threads exist	Fails until resolved (not yet a required context)
degraded verdict	Fewer than 2 models succeeded (infra)	Fails this job with a named reason

Powered by omniintelligence.review_pairing.cli_review — multi-model adversarial review: qwen3-review (Qwen3.8-27B), gpt-oss-review (gpt-oss-120b) (OMN-8468/OMN-8524/OMN-17492)

[jonahgabriel]

Lab: host=h201 lane=pricing-one-2a21 head=b1aa250bfde20a8ebce459531685ea2d34e248cf command="uv run pytest tests/ci/test_pricing_manifest_schema.py tests/unit/configs tests/unit/models/pricing tests/unit/nodes/node_savings_estimation_compute -q -n 0" observed="157 passed"; scripts/lint_pricing_manifest.py exit 0

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: qwen3-review, gpt-oss-review
Models failed: none
New finding threads: 4
Deduped (already posted on this PR): 0
Below quorum (one model only, reported not threaded): 6
Nit-level findings suppressed: 0

The model is the FINDER, never the gate: merge is gated only by the
deterministic Hostile Review Thread Gate, which blocks while
hostile-reviewer threads are unresolved. Resolve each thread after
addressing (or rejecting, with a reply) its finding.

Below quorum: 6 finding(s) raised by one model only (OMN-18479)

These are reported and NOT dropped, but they get no thread and do not block: a single model's finding no other model reproduced is not evidence enough to stop a merge. Read them; act on them if they are right.

• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — Floating-point equality assertion on YAML-parsed values is fragile | The test asserts entry.input_cost_per_1k == pytest.approx(input_cost_per_1k) after loading via ModelPricingTable.from_yaml(). I
• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — Test does not verify that local models are absent from paid-model lookup paths | The test confirms local models have zero cost, but does not verify that the pricing table does not accidentally treat t
• [MAJOR] src/omnibase_infra/configs/pricing_manifest.yaml (qwen3-review) — Model ID casing inconsistency between qwen3.8 and Qwen3.8-27B | The manifest defines both qwen3.8 (lowercase) and Qwen3.8-27B (uppercase Q). If the pricing table lookup is case-sensitive, these ar
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — No test verifies that unknown models raise or return a sentinel, not a silent default | The test suite validates that specific models are present and priced correctly, but does not test the behavior f
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (qwen3-review) — Test reads the manifest file twice: once via ModelPricingTable, once via raw yaml.safe_load | Each test function calls ModelPricingTable.from_yaml() and then separately calls `yaml.safe_load(_DEFAUL
• [MINOR] src/omnibase_infra/configs/pricing_manifest.yaml (gpt-oss-review) — External source URLs are not validated for safety | The manifest now includes several HTTPS URLs from third‑party providers. The code that consumes these URLs does not perform any validation or saniti

Findings demoted from threads (anchor rejected)

• [MINOR] hostile-reviewer (qwen3-review)

  Duplicated evidence block structure across all entries is a maintenance hazard | Every new entry repeats the same evidence sub-structure with identical generated_at, sample_window_days, sample_count, min_samples, query, and authoritative fields. This is a data duplication that will drift over time. If the query string or min_samples threshold changes, all 13 entries must be updated in lockstep. A schema-level default or a shared evidence template would reduce the surface area for inconsistency

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MINOR] hostile-reviewer (qwen3-review)

  estimate_cost test uses 1000 input and 1000 output tokens, but does not verify the formula | The test asserts estimate.estimated_cost_usd == pytest.approx(0.006) for 1000 input and 1000 output tokens at $0.001/1k input and $0.005/1k output. The expected value is 1000/1000 * 0.001 + 1000/1000 * 0.005 = 0.001 + 0.005 = 0.006. This is correct. However, the test does not verify the formula for non-1000-token counts. If estimate_cost has an off-by-one in the token-to-1k conversion (e.g., using integer divisi

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MINOR] hostile-reviewer (gpt-oss-review)

  Local model entries omit source_url in evidence block | The newly added local model entries provide an evidence section that lacks a "source_url" field, whereas paid model entries include this field. Downstream code that assumes the presence of a URL may raise a KeyError or skip validation, breaking the manifest processing pipeline. | Evidence: Local model entry for "qwen3.8" includes evidence with keys generated_at, sample_window_days, sample_count, min_samples, query, authoritative but no "source_url". |

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MINOR] hostile-reviewer (gpt-oss-review)

  Test reliance on default manifest path may be brittle | The new unit tests call ModelPricingTable.from_yaml() which reads the manifest from a hard‑coded default path. If the test runner's working directory differs from the repository root, the path resolution could fail, causing false negatives in CI environments. | Evidence: tests/unit/configs/test_pricing_manifest_fill_omn20372.py imports _DEFAULT_MANIFEST_PATH and uses it without adjusting for relative execution context. | Fix: Modify the test to locate

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: qwen3-review, gpt-oss-review
Models failed: none
New finding threads: 0
Deduped (already posted on this PR): 0
Below quorum (one model only, reported not threaded): 9
Nit-level findings suppressed: 2

The model is the FINDER, never the gate: merge is gated only by the
deterministic Hostile Review Thread Gate, which blocks while
hostile-reviewer threads are unresolved. Resolve each thread after
addressing (or rejecting, with a reply) its finding.

Below quorum: 9 finding(s) raised by one model only (OMN-18479)

These are reported and NOT dropped, but they get no thread and do not block: a single model's finding no other model reproduced is not evidence enough to stop a merge. Read them; act on them if they are right.

• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py:48-49 (qwen3-review) — Floating-point equality assertion on parsed YAML values | The test test_paid_models_have_documented_rates asserts entry.input_cost_per_1k == pytest.approx(input_cost_per_1k) where the expected value i
• [MAJOR] src/omnibase_infra/configs/pricing_manifest.yaml:430-460 (qwen3-review) — No test for model ID case-sensitivity or normalization | The manifest contains both 'qwen3.8' (lowercase) and 'Qwen3.8-27B' (mixed case) as distinct keys. The test parametrizes both and asserts they r
• [MAJOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py:60-64 (qwen3-review) — estimate_cost test does not verify input/output token attribution | test_haiku_cost_estimate_uses_the_documented_rate calls table.estimate_cost('claude-haiku-4-5', 1000, 1000) and asserts the total is
• [MINOR] src/omnibase_infra/configs/pricing_manifest.yaml:384-395 (qwen3-review) — Duplicate sample_count field at two nesting levels | Each new manifest entry has sample_count: 0 at the top level of the model entry and again inside evidence.sample_count: 0. The test only checks the
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py:33 (qwen3-review) — No validation that source_url is reachable or well-formed beyond https prefix | The test asserts raw_entry['evidence']['source_url'].startswith('https://') but does not validate that the URL is a well
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py:25-35 (qwen3-review) — Test loads and parses the full YAML manifest on every parametrized invocation | Each parametrized test case calls ModelPricingTable.from_yaml() and yaml.safe_load(_DEFAULT_MANIFEST_PATH.read_text(...)
• [MINOR] src/omnibase_infra/configs/pricing_manifest.yaml (gpt-oss-review) — Low confidence entries added without validation | The diff adds multiple model pricing entries marked with confidence "LOW_CONFIDENCE" and source "FALLBACK_PROVIDER_DOCUMENTATION". Downstream componen
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (gpt-oss-review) — Test suite loads full manifest for each parametrized case | Each parametrized test reads the entire pricing manifest from disk, causing repeated I/O and YAML parsing, which can degrade test suite perf
• [MINOR] tests/unit/configs/test_pricing_manifest_fill_omn20372.py (gpt-oss-review) — Assumption that retrieved_at equals effective_date may be brittle | The test asserts that "retrieved_at" matches "effective_date" for each new entry. If future updates record a different retrieval dat