Repository navigation
Releases
Open HI 0.23.1
Open HI 0.23.1 (Alpha: macOS Apple Silicon, Windows ARM64, Windows x64)
Same code as 0.23.0, which shipped for Windows ARM64 only, now on all three platforms.
- Each run is titled by the opening sentence of its delegation, kept only on the machine.
- A new run timeline: the flow of the run, with a details panel.
- The metrics screen is rebuilt in four scenes: Overview, Rhythm, Models and Work.
write_tolimits where a delegated session may write, for one file or one folder, and the receipt names every refusal.- The local worker no longer piles up old progress ledgers in a small context window.
This is an Alpha preview: the installers are not code-signed or notarized.
Open HI 0.23.0
Open HI 0.23.0 (Alpha, Windows ARM64)
- Each run is titled by the opening sentence of its delegation, kept only on the machine.
- A new run timeline: the flow of the run, with a details panel.
- The metrics screen is rebuilt in four scenes: Overview, Rhythm, Models and Work.
write_tolimits where a delegated session may write, for one file or one folder, and the receipt names every refusal.- The local worker no longer piles up old progress ledgers in a small context window.
This is an Alpha preview: the installer is not code-signed.
Open HI 0.22.0
Open HI 0.22.0 Alpha
YHITL Forjal is now Open HI, the first product from Forjal. Keep your cloud agent; Open HI adds a local one that works on your computer when your agent calls it.
What's new
- New name and look. Paper background, white cards and the Open HI mark across setup, Home, Models, Activity and Connections. Setup screens fit without scrolling.
- Reading by scope. Your agent can hand
run_local_readone glob or folder. The local model reads it one file at a time and returns one verified line per file withpath:line. An excerpt that is not in the file is barred and the file is named. - Shorter direction for the agent. The bridge names the loaded model up front and gives one short rule: hand off what would bring much more text into context than the answer; keep what a search answers or what must be read to edit.
- Claude Code rule file. Connecting Claude Code with the read consent installs
~/.claude/rules/forjal.md; disconnecting removes it. - Cancelled runs. A run stopped before it finished shows "Cancelled" in Activity and never counts as a failure.
- Clearer setup: one connection state everywhere, "Open a new session" instead of "Restart required", lanes named correctly, and the memory a model needs shown in the picker.
Upgrading
Existing installs update in place and keep their models, preferences and agent connections. Technical names (the forjal MCP server, the data folder) did not change.
Alpha builds carry no public signing identity, so macOS and Windows ask you to confirm the first launch.
YHITL Forjal 0.21.0
YHITL Forjal 0.21.0 Alpha
New
- Test a model in the harness. Open a model, go to Test, and give it a small job. Each prompt runs through the Forjal harness in a new, empty folder with write access, and the screen shows every step: each tool call, the model thinking, the generation, the answer, and the files it created. Prompt ideas follow the model's size.
- LFM2.5 2.6B, the smallest model in the catalog, on Apple Silicon (llama.cpp Metal, 16 GB+) and Snapdragon X (GenieX hybrid).
- LFM2.5 24B-A2B (24B total, 2.45B active) on Apple Silicon with 24 GB+ and on Snapdragon X with GenieX hybrid, where it now leads the max-capability slot.
- A new look. Onboarding and the workspace share one visual language: white paper, near-black ink and the model skies as the only colour.
Fixed
- The model picker showed the same few models in every Fast-to-Smart position. All published profiles now reach the app, and the short list features one configuration per base model.
- Claude Code installed with nvm, fnm, Volta, asdf, mise, pnpm, Yarn or a custom npm prefix, or through the Claude desktop app, is now found and runs from the app.
Notes
- Both LFM2.5 models are candidates: no quality, speed or token-savings claim is made.
- This is an Alpha build: installers are not signed or notarized.
YHITL Forjal 0.20.0
Added
- Forja V1 runs delegated workspace work (ADR-0064). On every llama.cpp deployment, a run granted
read,writeorrunnow runs in the Forja, the adaptive harness that configures itself from the deployment's datasheet. That covers Qwen3.8 27B and Qwen3.6 35B-A3B on macOS, and Ornith 1.5 9B and Qwen3.8 27B on Windows x64. The datasheet sets the tools, the descriptions, the prompt, the window and step caps and the loop checks for the deployment's size tier (small, medium or large). - Every runtime deployment ships with a Forja datasheet (132 in all).
- Each workspace result carries a
harnessfield. It names the harness, the Forja policy (forja.policy.v12), the tier, the configuration hash, and the levers whose setting is an estimate.
Changed
- The workspace tool loop now follows the datasheet. It used to be enabled only for a curated list, the Qwen3.8 Q3 Metal deployment alone. Now it reaches every llama.cpp deployment whose runtime parses tool calls. Natural worker discovery keeps its curated scope, and no model's candidate, compatible or certified status changed.
- The grant boundary is unchanged:
- the runtime keeps the roots and the credential deny-list;
- every command goes back to the runtime's allowlist executor, with no shell;
- the Forja worker can reach only the deployment's own loopback model server and cannot start processes;
- written files and executed commands are named in the result, also when a run fails.
- Answers keep the strict literal-evidence gate of the previous harness.
Notes
- The Forja's configuration on catalog models is estimated, not measured. No configuration was measured on the deployment it runs on, and every llama.cpp deployment in the catalog falls in the medium or large tier, whose doses are estimates (only the small tier repeats a setup measured on LFM2.5-2.6B). No quality, speed or token-savings claim is made.
- MLX, Snapdragon (GenieX), AMD (Lemonade) and Intel (OpenVINO) deployments have datasheets but no Forja session, because their runtimes do not parse tool calls. They keep their current behavior. Windows ARM64 has no llama.cpp lane.
- A run interrupted by a runtime restart during its Forja session fails instead of repeating, so no write or command runs twice.
- Software-verified: the Forja, runtime, MCP and desktop test suites pass, and a packaging smoke starts the worker from the signed macOS app. It has not been run with a loaded model on Apple Silicon yet, and the worker's Node mode on Windows has not been run on a Windows machine.
YHITL Forjal 0.19.0
[0.19.0] - 2026-10-01
Added
- Local map (ADR-0059): a deterministic, owner-only index of each granted project and a one-call evidence packet of ranked verbatim spans with
path:line, hash-checked on disk, answered without the runtime or a model. query_local_mapMCP tool, always loaded for Claude Code, plusforjal-codex query|explain|showand a standaloneforjalCLI.- Background enrichment by an explicitly chosen local model (
forjal-codex map enrich|collect), labelled as inferred and used only for ranking. - A
local-map-v1bench contract profile and a reproducible retrieval diagnostic.
Changed
- The resident MCP doctrine now puts the local map before search/read loops; the read worker stays explicit. The resident surface shrank from 8,184 to 8,175 characters.
- Claude and Codex prompt hooks announce the map once per session instead of relying only on curated worker cards.
- The managed Codex execpolicy allows the three read-only map subcommands; building, enriching and dropping a map still ask for approval.
- Onboarding follows the forjal.com landing: gray and slab cards with the landing's radii, regular-weight headings, and labelled pill buttons. It also fits windows down to the 860 x 680 minimum.
- Model surfaces use the landing's pastel skies instead of the red backgrounds, with ink text; the five backgrounds shrank from 12.4 MB to 224 KB.
- The model picker is ordered for the machine it runs on (ADR-0060): the app identifies the Mac by model ("13-inch MacBook Air", "Mac Studio") with its chip, cores and published memory bandwidth, estimates decode speed per quantization, and weighs it against the published Artificial Analysis index and the precision with a Fast-to-Smart control. Estimates and third-party scores are labelled as such; NPU lanes keep their curated order and runtimes are unchanged.
- On a Windows PC with a discrete graphics card (GeForce RTX, Radeon RX, Arc), the app reads the card's name and video memory from its driver and estimates GPU deployments that fit in that memory at the card's published bandwidth, instead of the system memory's. The hardware step names the card.
- Eleven new Apple MLX candidates: Qwen3.5 2B, the first model an 8 GB Mac can run; 8-bit variants of Qwen3.5 4B/9B, Gemma 4 E4B/12B/26B-A4B/31B and Qwen3.6 27B/35B-A3B for Macs with room for them; Gemma 4 E2B; and Qwen3.5 122B-A10B for 96 GB+. Each ships with a draft capability profile. No existing machine's slot picks changed.
Notes
- Software-verified only; not yet installed or evaluated on the test Mac. No session token-savings claim is made.
- Speed figures in the model picker are bandwidth estimates, checked only against one measured run (Qwen3.8 27B 4-bit on an M2). The 8 GB cohort has not been observed.
- Hardware tables and the ranking policy derive from Magnitude (Apache-2.0); see THIRD_PARTY_NOTICES.md.
YHITL Forjal 0.18.6
Forjal 0.18.6 — Qwen3.8 and clearer local runs
This Alpha consolidates the general workspace harness and makes live and historical runs easier to inspect.
- Qwen3.8 27B in Models. The interface shows the resolved deployment and quantization. On Apple Silicon with 24–31 GB of unified memory, explicit selection prepares the normal UD-Q3_K_XL candidate with llama.cpp / Metal and an 8K context. Weights are downloaded separately and verified. Other hardware cohorts retain their existing deployment selection.
- General workspace harness. Eligible native workspace runs use the model-directed tool loop, bounded navigation, command guidance and structured final evidence. Read, write and run capabilities follow the caller's grant. Recovery preserves the admitted policy; unsupported paths retain their existing contract.
- Better run visibility. See the current stage, elapsed time, completed turns, local actions and recorded output tokens. Expand details for prompt, cache and decode counters when available. The timeline distinguishes reads, writes, commands and supervisor tool calls, preserves generation errors, and separates execution completion from collection by the supervisor. A refresh failure keeps the last recorded steps available.
Qwen3.8 and the general harness remain supervised candidates, not certified capabilities. Workspace tools for Qwen3.8 are enabled only on the specified Metal Q3 deployment. The supervisor must review results, evidence, edits and command receipts. This release does not establish a session token-savings percentage or a speed or quality advantage over Qwen3.6.
Alpha distribution: the macOS application is ad-hoc signed, without Developer ID signing or notarization. Windows installers are unsigned. Operating-system launch warnings are expected. Choose the installer for your platform and verify its published checksum.
YHITL Forjal 0.16.0
Forjal 0.16.0 — Ornith 1.5 candidate on Apple Silicon
This release updates the macOS Apple Silicon build. The Windows ARM64 and
Windows x64 downloads remain on version 0.15.0.
Ornith 1.5 9B candidate
Forjal adds ornith-1.5-9b as a text-only candidate for eligible machines with
at least 16 GB of unified memory. After the user or supervising agent explicitly
selects that model_id, the load policy resolves one pinned Apple MLX deployment:
- 16–23 GB: official MLX 4-bit artifact;
- 24–31 GB: official MLX 6-bit artifact;
- 32 GB or more: official MLX 8-bit artifact.
The initial plans use an 8K context, thinking enabled, and MTP disabled. The
resolved deployment is prepared and activated atomically. A failure does not
silently switch quantization, runtime, context, offload, or model.
Prepared-plan fidelity
Prepared deployments now retain the adapter's pinned chat template and sampling
configuration in the immutable execution plan. Existing model profiles and
deployments remain available.
Availability and limits
Ornith 1.5 is a candidate. It is not compatible, certified, or recommended,
and this build has not completed end-to-end validation on physical Apple Silicon
hardware. Model weights are not bundled in the installer; the first load
downloads the selected artifact from its pinned upstream revision.
Alpha build. The macOS application is ad-hoc signed and not notarized, so
macOS asks the user to confirm the first launch.
YHITL Forjal 0.15.0
More local agents
Forjal adds managed adapters for Pi, OpenCode, OpenClaw, and Hermes Agent
alongside Codex and Claude Code. Each adapter follows the host's own
configuration format, timeout, ownership, and reload behavior. The four new
adapters currently have contract, configuration, and fake-host evidence;
real-host end-to-end validation is still pending.
Prepared deployments
Loading an explicitly selected model_id can now resolve one pinned deployment
for the detected hardware and prepare, verify, calibrate, and activate it as one
flow. A failure does not silently switch backend, quantization, context, runtime,
or model.
New candidate path
Qwen3.8 27B and the llama.cpp multi-runtime path enter as candidates on eligible
hardware. They are not certified or recommended yet. Qwen3.8 is excluded from
8 GB systems and Windows ARM64, and availability depends on the deployment
actually packaged for that platform.
Local operational metrics
The desktop adds local operational metrics and an opt-in shareable receipt.
Prompts, outputs, source code, paths, and URLs are excluded.
Alpha build. The installers carry no public signing identity. Windows may
show SmartScreen, and the macOS build is ad-hoc signed and not notarized.
YHITL Forjal 0.14.0
The local model works here now
Forjal used to be a bridge: the local model asked, your cloud agent executed, and a
hook carried the payload back. 120 measured sessions on a MacBook Air M4 said that
shape cannot pay for itself. Every action the local model needs from the parent costs
three of the parent's turns — one to receive the request, one to run it, one to
resume — and a turn re-reads the entire context. Measured against an agent that just
did the work itself: US$ 0.26–0.77 delegating, US$ 0.077–0.11 not.
The one path that won money was the opposite one, where the local model acts on its
own. So that is the product now.
workspace replaces local_read, with three grades ordered by what can change on
your disk:
read— search and read the directories your client declared. Nothing changes.write— also create files. Every one of them is named in the result.run— also run this project's build, tests, linters and read-onlygit. Build and
test output is the largest payload in real work, and it used to be the one thing
only the parent could fetch.
run contains write in fact and not by convention: no test suite runs with a promise
that it leaves your tree untouched.
run_command is not a sandbox and does not claim to be. A project's own test suite
is arbitrary code by construction. What is enforced instead is narrow and checkable:
the model never names a program — only one from an allowlist Forjal wrote, and within
it only the operations Forjal listed; there is no shell, so a pipe or a redirect is a
literal character to the program that receives it; every argument that could be a path
is resolved inside your granted directories and refused against the credential
deny-list; the environment is scrubbed of anything that looks like a secret and of
every interpreter hook before the child process exists; and every command that ran comes
back in the result with its exit status.
Three things the same measurement asked for
min_capability — you name the least capable deployment allowed to answer, and a
machine that cannot serve it refuses instead of trying something smaller. The same
delegation was measured delivering 24 of 24 sections twice and nothing once, purely
because the machine was busy enough for the selection policy to fall to a smaller model.
The policy is right and has not changed; what was missing was your ability to say a weak
answer is worse than none.
shortfall — an answer that is empty, that says it could not do the task, or that
names almost none of the documents it was given comes back marked. The output is still
returned whole; you are just told what it is.
evidence — an answer built on files comes back with the verbatim passage behind
each claim, and Forjal reads each one off the same disk after the answer was written.
A quote that is not really there stays in the list, marked. In 11 of 13 measured
sessions the agent re-read the material after delegating and paid for both paths;
forbidden to re-read it got 60–63% cheaper and was wrong every time, so the re-reading
was the correction it had to apply. This is what removes it.
Fixed
- Two delegations with the same prompt over different files no longer collide. A run's
identity left the files out, so six documents delegated with one prompt collapsed and
five callers received an answer about a document they had not asked about. If you are
on 0.13.0, this is the reason to update. - The Claude Code hook matches again. It compared tool arguments by equality, and
Claude Code attaches adescriptionto everyBashcall it runs — so the two never
matched, the hook stayed silent, and the payload travelled through your context anyway. - Malformed tool calls are repaired before they are refused. Single quotes, a trailing
comma, a code fence, a missing brace. Five unusable calls in a row ended one measured run
with nothing delegated; the intent was right every time. - The read budget now follows the context window of the deployment that will answer,
instead of a flat ceiling that ended searches with the window still nearly empty — and
the model is told what it has left.
Windows catches up
Windows ARM64 and x64 were still on 0.9.0. Both are on 0.14.0 with this release.
What is not measured yet
Everything above is built and tested — 483 tests across the runtime and the bridge — but
the harness itself has not been through a paired measurement on real work. The numbers
quoted here are from the sessions that produced the design, not from the code that came
out of it. Treat workspace: "run" as what it is: a new surface, in an Alpha.
Alpha build. The installers are native to each platform but carry no public signing
identity, so macOS and Windows both ask you to confirm the first launch. Each package
carries a private Python, the local runtime and the MCP bridge. Models are downloaded on
first use and are not bundled.