Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions .github/workflows/ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,8 @@ jobs:
run: spec/validate.sh
- name: Validate ECEB-0.1
run: scripts/eceb-conformance.sh
- name: Replay public ECEB study
run: scripts/eceb-study-check.sh
- name: Build documentation
run: cargo doc --all-features --no-deps --locked
- name: Package
Expand Down
22 changes: 12 additions & 10 deletions .impeccable.md
Original file line number Diff line number Diff line change
Expand Up @@ -17,22 +17,24 @@ showy. Confidence comes from inspectable evidence and honest boundaries.
### Aesthetic Direction

A bright standards-laboratory editorial: warm paper-white surfaces, ink-dark
type, a rare vermilion signal color, hairline rules, asymmetric technical
composition, and one evolving geometric glyph. Motion should reveal compilation
and verification as a sequence, carrying the visitor toward the specification
and repository links. Avoid dark cyberpunk, glowing AI gradients, generic SaaS
cards, glassmorphism, and decorative code rain.
type, a rare vermilion signal color, hairline rules, and asymmetric technical
composition. Motion should reveal corpus → roots → source-verified context as a
sequence, carrying the visitor toward the protocol, public evidence, and
repository. The geometric mark is a brand signature and optional fingerprint,
not the context mechanism. Avoid dark cyberpunk, glowing AI gradients, generic
SaaS cards, glassmorphism, and decorative code rain.

The display face is Familjen Grotesk and the reading face is Public Sans. The
combination should resemble a carefully typeset protocol exhibit rather than a
developer dashboard.

### Design Principles

1. Show the mechanism before asking for belief.
2. Make evidence and limitations equally visible.
3. Use motion to explain state transitions, never as ambient noise.
4. Keep the page fast, dependency-light, keyboard-usable, and respectful of
1. Lead with the research protocol and source-verifiable context compiler.
2. Keep protocol guarantees, measured findings, and future hypotheses distinct.
3. Make evidence and limitations equally visible.
4. Use motion to explain state transitions, never as ambient noise.
5. Keep the page fast, dependency-light, keyboard-usable, and respectful of
reduced-motion preferences.
5. End every narrative path at an inspectable artifact: the specification,
6. End every narrative path at an inspectable artifact: the specification,
conformance command, dataset, or repository.
8 changes: 8 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,17 +14,25 @@ the project intends to use Semantic Versioning after its first stable release.
- Published semantic-routing gold seed and blind-discovery proof documentation.
- ECEB-0.1 draft specification, JSON Schemas, multi-source enterprise
conformance seed, executable evaluator, and negative policy/provenance gates.
- Deterministic 30-task Multi-SWE-bench Rust study suite, four-arm runner,
comparative study schemas, raw golden run records, and offline replay gate.

### Changed

- Manifest verification now checks the schema and all recomputed derived records
in addition to corpus and semantic roots.
- SVG artifacts declare the `symgliph.glyph/svg-v1` rendering profile.
- Public positioning now leads with the research protocol and source-verifiable
context compiler; the SVG is identified as an optional sibling profile rather
than the model-facing token-reduction mechanism.

### Experimental evidence

- The fixed Fabric blind-discovery pilot completed 18/18 exact selections while
reducing native input tokens by 85.86%. This is a scoped result, not a universal
performance claim.
- The Multi-SWE-bench Rust v0.1 study reduced paired prompt tokens by 97.58% and
localization-model spend by 98.54% with Symgliph compiled context, while exact
fix-path recall declined from 53.32% to 41.76%; lexical BM25 retained 46.87%.

[Unreleased]: https://github.com/copyleftdev/symgliph/commits/main
11 changes: 8 additions & 3 deletions Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -4,15 +4,15 @@ version = "0.1.0"
edition = "2024"
rust-version = "1.88"
default-run = "symgliph"
description = "Compile a corpus into a verifiable semantic glyph artifact"
description = "Research protocol and context compiler for source-verifiable AI context"
authors = ["copyleftdev"]
license = "MIT OR Apache-2.0"
repository = "https://github.com/copyleftdev/symgliph"
homepage = "https://symgliph.org/"
documentation = "https://docs.rs/symgliph"
readme = "README.md"
keywords = ["corpus", "merkle", "visualization", "provenance"]
categories = ["data-structures", "visualization"]
keywords = ["ai", "context-engineering", "provenance", "retrieval", "rust"]
categories = ["development-tools", "data-structures"]
exclude = [
"/.github/**",
"/.impeccable.md",
Expand Down Expand Up @@ -68,6 +68,11 @@ name = "symgliph-golden"
path = "src/bin/symgliph-golden.rs"
required-features = ["golden"]

[[bin]]
name = "symgliph-eceb-study"
path = "src/bin/symgliph-eceb-study.rs"
required-features = ["openrouter"]

[lints.rust]
unsafe_code = "forbid"
missing_docs = "warn"
Expand Down
Loading