docs(wiki): vf-clide page memory commands to v1.0.5 (catch-up)
The Memory section and the REPL command list were frozen at v1.0.3 and had
missed the v1.0.4 connection layer entirely. Catch them up to the current
surface, verified against the repl.rs parser and the Memory page: recall flags
(--explain / --type / --include-superseded / --frontier), /retype, the
SUPERSEDES / DERIVES_FROM / /why connection layer, and the v1.0.5 CONTRADICTS
edge (/contradict, /uncontradict). The agent tool set stayed three
(recall/remember/archive) across v1.0.4-v1.0.5 — the diagnostics, typing, and
edges are user curation, never agent tools. No invented commands; additive only.
fcba525
docs(wiki): Memory + Home pages to v1.0.5 — conflict edges, opt-in frontier, edge-type priors, determinism
Additive update from v1.0.4 (v1.0.3/v1.0.4 content preserved). The Memory page
gains the CONTRADICTS edge (/contradict · /uncontradict — symmetric,
awareness-only, resolved via /supersede), opt-in frontier retrieval (--frontier,
off by default), edge-type priors (CONTRADICTS as a frontier negative signal),
and cross-process recall determinism (VF_HNSW_SEED on SQLiteGraph 3.3.1). The
endpoint + command surface is verified against routes.rs, the REPL parser,
README, and api_server_architecture.md §5.8 — no invented commands. The
SQLiteGraph reference is updated to 3.3.1 (maeddesg fork). The Home banner is
bumped to v1.0.5 with a matching changelog entry. Invariants kept prominent:
opt-in, default recall byte-identical, edges never change ranking, the user
curates.
55ff20f
docs(wiki): Memory page to v1.0.4 — recall diagnostics, note typing, edges
Memory.md: document recall --explain, VF_RECALL_MARGIN (opt-in), note
typing (5 types + --type/retype/filter), SUPERSEDES (suppress-default,
chain, backfill-to-k, mark-not-destroy, --include-superseded),
DERIVES_FROM + /why (cycle/depth-guarded, never alters recall); extend
endpoint block + 404 note; correct roadmap. Home.md: v1.0.4 highlights
section + version note. _Footer.md: v1.0.3 -> v1.0.4.
bde4779
Document v1.0.3 — agent-side archive, un-archive, curation 404
Memory.md: add POST /memory/unarchive + the 404-on-missing-id note;
curation now reversible (agent may archive behind confirmation, /unarchive
and /forget user-only). Memory-Design.md: ground the "mark, don't destroy"
principle in the shipped /archive+/unarchive mechanism. vf-clide.md: third
memory tool (archive, confirm-always), /unarchive REPL command, curation
split. Home.md: v1.0.3 section + "documents v1.0.3". _Footer: v1.0.2 -> v1.0.3.
92532e1
Updated Memory (markdown)
ecc8e38
v1.0.2: vf-clide memory access + curation; Memory Design page; footer bump
9b7b0e3
docs(home): bump banner to v1.0.1
Home now documents the shipped v1.0.1 reality; the memory section heading
tracks the v1.0.1 opt-in change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
7cc4fae
docs(memory): document opt-in reality across the wiki
Bring the wiki onto the opt-in memory (commit 35bbec1: Cargo feature `memory`
default off + runtime flag `serve --memory`). Memory was missing or read as
always-on on seven pages.
- Memory.md: new "Enabling memory" section (two gates, paths, rustc 1.89 floor);
corrected "embedded in serve" to opt-in; subtitle notes off-by-default.
- Installation.md: prerequisites rustc floor (lean 1.85+, memory 1.89+),
`--features memory` build command, new "Memory (optional)" setup subsection.
- Usage.md: serve section gains the `--memory` enabling path, /memory/* curl
examples (remember/recall/projects), and the 503-without-flag note.
- Hardware-and-Compatibility.md: "Memory (optional) — CPU, not VRAM" (AVX-512/
VNNI fast path, MLAS fallback; no VRAM, doesn't touch the 16 GB budget).
- Configuration.md: "Memory (optional, default off)" flag table
(--memory / VULKANFORGE_MEMORY / VF_MEMORY_DB).
- Troubleshooting.md: memory FAQs (503, rebuild message, first-start download,
where it's stored).
- Home.md: v1.0 memory section reframed opt-in (two gates, cost only with feature).
Facts verified against the binary (serve --help, default paths) and declared
MSRVs (sqlitegraph rust-version 1.89, ort 1.88), not the brief. No Stufe-B docs
(REPL/agent memory tools don't exist yet). Goes out with the code release (mg).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
7aa3686
v1.0 — server-side memory: dedicated Memory page (philosophy + how-it-works + roadmap); Home banner + section; sidebar
e93eb46
v0.9.4 — vf-clide REPL honors the permission ceiling; clearer denials
67c10eb
v0.9.2 — vf-clide token meter + pinned status line; clean serve shutdown
5e423ac
docs(wiki): v0.9.1 — note search is workspace-confined (skips escaping symlinks)
Earlier 0.9.0/0.2.0 builds' search recursively followed symlinks out of
the workspace; v0.9.1 skips them. Update recommended.
cf2ecdd
docs(wiki): v0.9.0 — agentic vf-clide
Rewrite the vf-clide page from chat-only to chat + agentic: --agent tool
loop (read_file/write_file/search/shell), three-tier permission model
(--yes/--allow-mutating/--allow-shell, cumulative), workspace confinement,
constitution (default system prompt + AGENTS.md), and the honest limits
(shell not confined → --allow-shell; no persistence; substring search;
16384 ctx; gemma simple args). Add a v0.9.0 Home section and update the
"chat client" descriptors on Home/Installation/Usage.
cec097a
docs(wiki): v0.8.0 — auto ctx-size, Gemma-4 tool-calling, vf-clide; KV-FP8 now required
- New page: vf-clide (the standalone CLI chat client) + Sidebar/Home/Usage/Installation links.
- Automatic context sizing (v0.8.0): new Configuration section, Usage/Home/Troubleshooting notes,
the 16384 RDNA4 LDS ceiling (explicit --ctx-size above it aborts, not clamped).
- Gemma-4 native tool/function calling documented (Usage, Supported-Models).
- KV-FP8 corrected from "recommended" to REQUIRED for the Gemma-4-26B-A4B MoE across
Supported-Models / Configuration / Usage / Troubleshooting / Architecture / Choosing-a-Model:
the non-FP8 KV path is known-broken and the engine fail-loud aborts without VULKANFORGE_KV_FP8=1
(debug override VULKANFORGE_ALLOW_BROKEN_KV=1).
- Home/Configuration now reference the shipped v0.8.0 (perf matrix provenance stays v0.7.0).
69185ab
docs(wiki): add "Choosing a Model for Coding" comparison page (neutral, no default)
Side-by-side of the three coding-capable models (Gemma-4-26B-A4B Q3_K_M,
Gemma-4-26B-A4B QAT/Q4_0, Qwen3.6-27B Q3_K_S) by quality / speed / context,
from VulkanForge's own coding tests on the 16 GB RX 9070 XT. No single
"best" — user picks by priority. Honest caveats (small sample; not a
quant-controlled comparison; 16 GB-specific). Linked from _Sidebar,
Supported-Models, and Benchmarks; softened the stale Q3_K_M "default" note.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
d744b8f
docs(wiki): add VF_USE_GRAPH to Configuration
Opt-in graph-dispatch flag (default off; topologically-sorted per-layer dispatch vs the default
imperative loop). Effect + default sourced from the v0.7.0 code; neutral wording.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
6f0148d
docs(wiki): VulkanForge user wiki @ v0.7.0
9 content pages + sidebar + footer, sourced strictly from the v0.7.0-tagged repo (README,
CHANGELOG, docs/BENCHMARKS.md, docs/MODELS.md, Cargo.toml, LICENSE, real flag/CLI code):
Home, Installation, Hardware-and-Compatibility, Supported-Models, Usage, Configuration,
Benchmarks, Architecture, Troubleshooting.
User-facing, English, honest (v0.7.0 same-run matrix vs llama.cpp Vulkan: dense prefill parity
0.93-1.04x, Gemma-MoE @2048 0.83-0.89x, decode 0.87-0.97x). No pre-v0.7 numbers, no internal
(sprint reports / coding-standards / bisect) leaks; flags grepped from code with verified defaults.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
02792bd