Refs librefang/librefang#4842 — long-term replacement for the substring
match that the OpenAI driver currently uses to decide how to handle
`reasoning_content` on historical assistant turns.
Three provider-specific behaviours that the driver must distinguish at
wire time, now expressed as catalog metadata:
* `strip` — DeepSeek R1 / deepseek-reasoner. The API rejects requests
that carry reasoning_content on previous assistant messages.
* `echo` — DeepSeek V4 Flash. Thinking mode is on by default and the
API rejects multi-turn requests when assistant turns containing
tool_calls don't echo back the original reasoning text. This is the
bug surfaced in librefang/librefang#4842.
* `empty_string` — Moonshot / Kimi K2 family. The field must be present
(empty string) on tool_calls turns, with thinking disabled wire-side
for multi-turn compatibility.
* `none` (default) — most providers; field is omitted entirely.
V4 Pro is intentionally NOT marked `echo` — librefang#4842 reports it
working out-of-the-box; flip when there's an empirical reproducer.
Marks affected models:
providers/deepseek.toml
deepseek-v4-flash → echo
deepseek-reasoner → strip
providers/moonshot.toml
kimi-k2.6, kimi-k2.5, kimi-k2 → empty_string
providers/kimi-coding.toml
kimi-for-coding → empty_string
providers/byteplus-coding.toml
kimi-k2.5 → empty_string
providers/novita.toml
moonshotai/kimi-k2-thinking → empty_string
Tooling:
* schema.toml registers the field with the four enum options and a
`none` default so existing TOML files keep parsing unchanged.
* scripts/validate.py rejects unknown enum values; verified with a
hand-crafted negative case (`reasoning_echo_policy = "bogus"` →
validation fails with the expected message).
* `python3 scripts/validate.py` passes (267 models).
The librefang side that consumes this field will land in a follow-up
PR — until then, registry consumers ignore the field via
`#[serde(default)]` and the existing substring fallback continues to
work, so this commit is safe to ship independently.
Extend modality enum to support video and music, then register the
non-text MiniMax models that were already declared in
media_capabilities but had no concrete entries:
- image-01 ($0.0035/image)
- speech-2.8/2.6 hd & turbo ($60-$100 per 1M chars)
- Hailuo 2.3 Fast / 2.3 / 02 video models ($0.10-$0.56 per video)
- music-2.6, lyrics_generation
Per-call pricing is documented in inline comments since the schema's
token-based cost fields don't naturally fit per-call billing.
schema.toml and scripts/validate.py both updated; the change is
additive (existing modality values remain valid).
Introduces image-generation models as a first-class [[models]] entry via
a new `modality` field on the model schema ("text" default, "image",
"audio"). When modality != "text", context_window / max_output_tokens
are optional since no conventional context gate exists — OpenAI's
gpt-image-2 docs omit them.
Adds `image_input_cost_per_m` / `image_output_cost_per_m` alongside
existing text token cost fields to cover the 4-price structure OpenAI
uses for image generation (text $5/$10, image $8/$30 per 1M tokens).
Validator updated to:
- accept any modality in {text, image, audio}
- require context_window/max_output_tokens only for modality=text
- range-check the two new cost fields
gpt-image-2 entry added to providers/openai.toml with pricing sourced
from https://developers.openai.com/api/docs/pricing. Snapshot
gpt-image-2-2026-04-21 listed as alias.
Standardize on Claude Code's SKILL.md format as every skill's source of
truth. skill.toml becomes an optional metadata layer for runtime, input
schema, and versioning — never for the prompt body.
- validate.py: require SKILL.md in every skill dir; when skill.toml also
exists, cross-check name/description consistency to prevent drift
- Add SKILL.md to the two custom-skill examples
- Move the meeting-agenda prompt body out of skill.toml into SKILL.md
- Rewrite skills/README.md to document the md-first, toml-as-metadata convention
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude Code-style skills use SKILL.md with YAML frontmatter instead of
skill.toml. Validator now accepts either form, unblocking the 60 bundled
skills restored in #42.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(validate): check for [agents] instead of [agent] in HAND.toml
All 14 hands use [agents.main] (plural) for multi-agent config,
but the validator was checking for [agent] (singular), causing
all hands to fail validation.
* fix(routing): resolve 19 routing alias collisions
Agent is a sub-unit of hand, so hands take priority for routing.
Remove conflicting aliases from agent side when hand already owns them.
- analyst: remove data analysis, analyze data, dashboard (owned by hand/analytics)
- data-scientist: remove statistical analysis, forecast, prediction (owned by hand/analytics, hand/predictor)
- sales-assistant: remove prospecting, sales, pipeline (owned by hand/lead, hand/devops)
- devops-lead: remove incident response, kubernetes, terraform (owned by hand/devops)
- researcher: remove deep research, research, literature review (owned by hand/researcher)
- academic-researcher: remove literature review, systematic review (owned by hand/researcher)
- social-media: remove duplicate content calendar from weak_aliases
- hand/collector: remove competitive analysis (owned by hand/strategist)
Community-maintained TOML catalog for LibreFang. New models can be added
via PR without requiring a LibreFang binary release.
Includes validation script, bilingual docs, and GitHub templates.
Evan