RPG OS v0.9.4

RPG OS v0.9.4 — universal encounter generation

An encounter generator built around named human moods and venue-specific activities cannot serve an unfamiliar creature, machine or setting. v0.9.4 provides an optional reusable procedure: establish the entity's actual capacities and circumstances, obtain direction and intensity for eligible gaps, then have the GM assign concrete current states before dependent behavior.

What changes

The five-field temporary basis remains Condition and mode; Priorities and constraints; Perception and appraisal; Attraction; Engagement stance. The optional helper uses two independent d6 per eligible unfixed condition, initial interpersonal appraisal or overall attraction:

2d6 total	Direction and intensity
2–3	Strongly negative
4–5	Negative
6–8	Neutral or mixed
9–10	Positive
11–12	Strongly positive

At most three determinations use six independent faces. The GM fixes the eligible facet before input and fills a concrete compatible value after its result, before the entity acts. There is no six-mood catalogue, custom outcome list for each entity, or additional intensity/subtype roll. The middle band must establish a concrete state, not an empty neutral placeholder; strong values remain within supported capacities and create no compulsory response.

Priorities and constraints come from the entity's own actual aims, processes and circumstances. Perception stays factual. Nonsocial entities use their sourced sensing, control and response mechanisms. Engagement follows from the complete basis, without another willingness roll. Attraction remains distinct from trust, availability and participation.

ENGINE/_shared/ENCOUNTER_GENERATION.md owns the reusable method. MODULE binds real world and entity sources; INSTANCE retains current facts. Generic activation, retention and setup guidance uses actual participation rather than a particular venue. Prescribed first-focus portrayal retrieval, group scale, active-state reuse, genuine deactivation and consequential memory remain in place.

Adoption and compatibility

Freeform is the only bundled selectable engine. It can explicitly select this helper through the accepted agreement; ordinary Freeform judgment remains the default. The helper is not another engine or an automatic dice policy. It changes no other engine's success checks, existing active baseline, resolved outcome, save schema, PC decision or autosave permission.

Existing campaigns retain their accepted methods. Use the upgrade procedure for a program update and recalibration for an accepted prospective method change. Supported v0.9.3 and earlier compatible records retain their identities and formats. Private modules, people, campaign bindings, source assets and audit transcripts are excluded from the public kit.

Verification

New checks cover the optional helper, five bands across all 36 independent dice pairs, input limits, source ownership and preservation of the existing lifecycle. Both GitHub workflows run the encounter suite and verify the committed unbound package. See observed verification for actual results and limitations.

Five small prompted synthetic rehearsals exercised an adult, sourced predator, controlled gates, bacterium and missing-source entity; the rolled cases were repeated under the final 2d6 scheme. These are bounded observations, not proof of natural-play or long-context reliability. Published v0.9.3 tags and assets remain unchanged.

RPG OS v0.9.3

RPG OS v0.9.3 — temporary encounter state and consequence continuity

v0.9.3 gives individually participating entities a concrete current basis before their behavior, keeps that basis while needed, and releases inconsequential temporary values when participation reasonably ends. It also makes focused portrayal-source retrieval and two consequence-handling duties explicit in the existing operating cycle.

The public kit remains unbound and ships Freeform only. It includes no campaign world, character, live encounter state or test campaign.

Establish before behavior

Every PLAY response follows Understand → Establish → Resolve → Portray. The model identifies the actual declaration, recovers sufficient authority, resolves from the established basis, then presents the supported perceptible result. The stages can be brief for routine play. A module-prescribed first-focused-portrayal guide must be available before focused physical description; previously retrieved guidance may be reused within its coverage.

Before individual participation, appraisal, initiative or a separately resolved personal decision involving the PC, complete the Initial Encounter Baseline:

Current field	What it establishes
Condition and mode	Present physical, emotional or operational condition and immediate activity.
Priorities and constraints	Current aims, needs, commitments, orders and practical limits.
Perception and appraisal	Actual detection, received information and current interpretation, including mistakes.
Attraction	Where applicable, overall attraction, no particular attraction or aversion within the supported domain.
Engagement stance	Current openness and boundaries, or the appropriate nonsocial response policy.

Each field has a concrete value, supported inapplicability, or a precise source/perception limitation before dependent behavior. Nature, control, capabilities, identity and established history constrain the current state; activation does not require inventing enduring temperament or a biography. Applicable attraction is established even when a person is on duty or partnered. Its finer physical, aesthetic, sexual or romantic meaning may be resolved later when relevant; it does not determine trust, availability, participation or consent.

These are entity-appropriate determinations. A creature, organism, machine or collective uses its actual senses, drives, directives and control. Chemical response and predation do not become interpersonal attraction by analogy. Setup binds usable compatible methods, source routes, constraints and outcome meanings. The OS adds no universal human dice table, and five fields do not require five rolls.

Keep active state; release expired trivia

Merely observing an unaware individual does not activate private state. An individual's own look, appraisal or initiative can activate it before conversation. Focused physical portrayal follows its source trigger independently.

Coordinated conduct uses genuinely shared operational facts. Ten soldiers following patrol procedure need not receive ten personal baselines; a separately responding individual does. Three patrons individually deciding about offered drinks require three bases. Shared control does not grant shared knowledge beyond actual channels.

While participation continues, retain current values through pauses and shifts of attention and update affected fields for actual developments. Genuine deactivation permits releasing inconsequential temporary values. Preserve material observations, identity, relationships, promises, injuries and unresolved consequences. Physical absence alone does not end an ongoing causal exchange.

Later incidental contact may receive fresh eligible temporary values within surviving facts. Historical evidence remains evidence of its time; it does not automatically make expired values current again. A missing still-active or consequential fact is a source gap, not a permitted reset.

Authorized saves place the active temporary set in NOW and consequential individual facts in PEOPLE or their actual system owners. Scene handover preserves the same ownership and limits. This is a working-state policy: it neither erases earlier conversation tokens nor adds an inactive registry, permanent dossier for every contact, per-turn file write or background simulation.

Carry consequences forward

The two additions in LAW's Resolve stage require the GM to:

• Carry established event changes into current working state before continuing, including what affected entities actually perceived. Preserve observation limits and separate observations from interpretations. An event's actual reach governs its audience; unchanged bystanders need no extra entries.
• Check the current conditions, information, means and resources before executing a pending intention or prepared development. Resolve actual interruptions or supported alternatives, preserve unresolved matters, and require a real source and causal development for replacements, repairs or replenishment.

An intention is not a completed action. These duties use existing owners and do not authorize PLAY file writes.

Existing campaigns and evidence limits

Use the compatible upgrade procedure in a protected copy. Valid v0.7.3, v0.8.0, v0.8.1, v8.1.1, v0.9.0, v0.9.1 and v0.9.2 records retain their formats and save identities. Never install blank INSTANCE or ARCHIVE templates over a campaign. Preserve established people, histories, agreements and earlier determinations. Use the accepted method for eligible current gaps; settle only any missing prospective method or permission through upgrade or recalibration. Installation alone does not enable autosave or change a diceless agreement.

Limited paired crowded-scene trials included full saves and fresh-context continuations. They demonstrated successful retention and information-boundary handling in some tested cases, alongside disclosure defects and incomplete coverage. They did not demonstrate a clear advantage from the two Resolve additions or establish long-context memory reliability. Witnesses spoke publicly in both High-reasoning runs, so that repeat did not settle the silent-witness concern. Structural and instruction checks verify their stated scope; they cannot prove private baseline establishment, enjoyable play or model compliance. See verification for observed evidence and limits.

RPG OS v0.9.2

RPG OS v0.9.2 — optional announced autosave

This experimental release adds an opt-in, turn-driven autosave policy while preserving the existing campaign formats, state ownership, evidence boundaries and protected-write procedure. The policy invokes the ordinary complete-present CHECKPOINT; it does not add a background service, a second save format or an invisible private memory store.

Announce, checkpoint, verify

With the policy explicitly accepted, a meaningful unsaved change or substantial scene boundary can schedule a checkpoint. The standard backstops are 15 completed PLAY replies since verified persistence and a supported context reading of at least 65%. Context readings must come from actual model-visible host telemetry or an explicit operator report, never an estimate or an assumed ability to read the UI.

The GM announces the checkpoint one response before automatic execution. Several triggers coalesce into one pending notice. After the next completed play response, it enters ADMIN, preserves the complete present through the existing recovery-protected CHECKPOINT, verifies the writes, and reports the save identity and evidence limitation briefly. Nothing runs while the chat is idle. The pending PC choice and fictional time remain unchanged.

The player can request immediate CHECKPOINT or full SAVE, delay/resume autosave, or disable it. A manual full save is never downgraded. Recovery, active scene handover and an in-progress write prevent concurrent autosave. Failure preserves the recovery marker and does not claim success or reset the pending operation.

Save before switching

The outgoing context must save and verify before a planned model, reasoning-setting, provider, chat or GM change. The warning does not encourage switching to a stronger model before saving. This is a conservative continuity rule, not a promise about any host's internal context-transfer behavior.

A checkpoint protects current state but does not archive new dialogue. Use a full CLOSE or the existing scene-handover procedure before discarding source access; disclose the evidence gap if the operator elects checkpoint-only transfer. The incoming GM reloads authoritative saved records rather than reconstructing missing hidden facts or rolls.

Integration and verification scope

• ADMIN/AUTOSAVE.md defines selection, triggers, notice, execution, overrides, failures and transitions.
• LAW, BOOTSTRAP, NEW_GAME, RECALIBRATE and CLOSE_CONTRACT route the policy through the existing permission and persistence boundaries.
• TOOLS/autosave.py is an optional deterministic, read-only scheduling helper. It evaluates supplied observations; it does not inspect the model, read campaign state, write files or execute saves.
• TOOLS/test_autosave.py covers scheduler decisions, notice ordering, overrides, source-labelled telemetry, blocked operations, route integration and synthetic checkpoint/evidence metadata.
• ADMIN/TEST_AUTOSAVE.md contains ten separately unrun live behavioral cases. Automated tests do not establish model compliance, complete hidden-state retention or cross-host reliability.
• Both repository workflows run the new suite alongside the existing structural, handover, packaging, exact-source, search, evidence and agent-instruction suites. Release packaging still checks the exact committed tree and SHA-256.

See the release's GitHub Actions runs for the actual tested commit, platform skips and publication result. Do not interpret this release note as a claim that a future run has passed.

Compatibility

Autosave is off unless selected for that campaign. Installing v0.9.2 alone does not enable it or edit a separately stored campaign. Existing v0.7.3, v0.8.0, v0.8.1, v8.1.1, v0.9.0 and v0.9.1 record formats remain compatible through the protected upgrade procedure. Accepted agreements, private facts, individual/controller state, PC control, real-randomness requirements and the v0.9.1 prospective-deepening rule remain intact.

There is no new mandatory agreement clause, metadata field, commit kind, personality vector, autosave file or startup test suite. Ordinary PLAY stays read-only; accepted autosave enters ADMIN for the actual writes. Repeated checkpoints do not replace full CLOSE or independent backups.

To select the standard policy in a campaign, say: “Enable the standard announced autosave policy in ADMIN/AUTOSAVE.md for this campaign.”

Quick start · Autosave procedure · Behavioral cases · Installation and upgrades · Verification

RPG OS v0.9.1

RPG OS v0.9.1 — universal agent basis and prospective deepening

This experimental release refines independent-agent establishment so an NPC, creature, machine or collective is portrayed from its own supported nature, source of control, applicable drives/directives, information and circumstances rather than from the player character's interpretation or a universal human personality template.

It also closes a residual late-deepening loophole found during a synthetic playtest: once outward behavior has been portrayed, a later private detail must not be invented as the hidden pre-existing cause of that behavior merely because it fits the player's interpretation, or because it contradicts that interpretation and therefore appears "independent." Unexplained prior causes remain unresolved unless recovered from an existing authority or independently established through an accepted procedure using admissible inputs.

Added and changed

• Entity-appropriate behavioral basis: independence from the player is distinct from individual autonomy. People, creatures, machines, externally controlled bodies and collectives use only dimensions that actually apply to them.
• Control and information separation: shared control does not imply shared knowledge. Local perception, controller knowledge, communication channels, delays and local discretion remain distinct.
• Interpretation boundary: the PC's unverified interpretation, an unperceived player wish, a proposed narrative role and a desired success or refusal are not evidence of private nature. Actual observations constrain possible private facts without automatically establishing their perceived meaning.
• Inapplicable dimensions stay inapplicable: a relationship, emotion, social capability or other dimension is not treated as an unknown or zero score merely because a human character would have it.
• Prospective deepening: new evidence may change current beliefs, aims, orders or orientation from that point forward. It does not rewrite that change backward into an earlier motive or hidden history.
• Past-cause protection: an earlier private cause may be established later only from an already authoritative source or an accepted generation method whose admissible inputs do not include the player's interpretation or a need to explain the old portrayal. If several causes remain compatible and no independent determination is required, the cause stays unresolved.
• No contrarian laundering: choosing the opposite of the player's interpretation is not proof of independence and cannot be used as a hidden-cause generator.
• Sparse ownership preserved: no personality spreadsheet, controller schema, cast-wide simulation, mandatory startup read, private scratch layer or automatic save is introduced. Shared state keeps one existing owner.
• Compatibility preserved: existing characters, controllers, relationships, history and generation results are not regenerated or relabeled when adopting v0.9.1.

Regression coverage

TOOLS/test_agent_state.py now checks the universal input boundary, control/information separation, inapplicable dimensions, sparse ownership, persistence limits and the prospective-deepening rule. ADMIN/TEST_AGENT_STATE.md contains fifteen synthetic behavioral cases, including the new U15 case for an observed presentation followed by a player interpretation and a later attempted backstory explanation.

The behavioral cases remain evidence protocols rather than automatic LLM simulations. Passing text/route checks does not prove unbiased generation or hidden pre-establishment. A claimed private determination is not proof that it existed earlier.

The existing structural, handover, packaging, exact-source, scoped-search and evidence suites remain part of CI, together with the agent-instruction suite.

Preserved

v0.9.1 keeps the v0.9.0 save, agreement, archive, person/system and handover formats. The standing d6 fallback remains optional and unchanged. Existing source-access, search and evidence-audit tools remain available. Ordinary PLAY remains read-only; unsaved private state is best-effort until an authorized save or checkpoint succeeds.

Compatible v0.7.3, v0.8.0, v0.8.1, v8.1.1 and v0.9.0 campaigns can adopt the program update in a protected separate copy through ADMIN/UPGRADE_V08.md. The update itself does not change an existing agreement, engine selection, campaign facts or generation permissions.

Use it

Quick start · Installation · Independent agent state · Targeted agent cases · Verification

This remains an experimental release. Structural and instruction checks do not establish long-campaign, cross-model or hidden-state reliability.

RPG OS v8.1.1

RPG OS v8.1.1 — individual baselines on direct attention

This incremental instruction update follows v0.8.1. It establishes a small current individual basis before portraying a directly attended NPC or resolving their reception. Current campaign formats and resolution-engine rules remain unchanged.

Download RPG_OS_v8.1.1.zip and its SHA256SUMS from this release. The public package is a separate unbound fresh install with Freeform and empty campaign/archive records; it includes no private campaign, character, transcript or artwork.

What changes in play

“I approach her” is enough. When the player singles out an individual for direct attention or interaction, or an NPC directly engages the PC, the GM retrieves or establishes a minimal current baseline before focused portrayal or reception. It does not assume the PC wants company, directions, flirtation or confrontation. Quiet observation supplies neither an approach nor automatic NPC awareness.

A few relevant facts about current activity, orientation, manner or constraints can suffice. Existing people retain their established state. Even a brief clerk or passerby interaction can have a tiny baseline, while background crowds need no individual initialization. Deeper facts and durable records follow actual developments and continuity needs, without a mandatory personal-life inventory.

The NPC responds to what it can actually perceive or legitimately know. Its interpretation may be mistaken and does not declare the PC's intent. Applicable ENGINE procedures still resolve uncertainty; individual description does not invent numerical modifiers. Initial state, reception and later willingness retain their separate scopes. Do not resolve the same question twice or invent a past motive after seeing the outcome.

The standing d6 fallback remains available for eligible unresolved outcomes with too little basis for grounded judgment. Facts, applicable rules and other selected methods keep priority. Existing diceless agreements remain intact, and standing selection still needs no per-roll permission.

Continue an existing campaign

Use the compatible-campaign upgrade procedure in a separate protected copy. Preserve the old runtime, campaign records and evidence. Do not extract blank INSTANCE or ARCHIVE templates over a played campaign. The update does not reset a run, revise its agreement, initialize past encounters or reroll existing people.

PLAY remains read-only. New unsaved state has best-effort retention in the available conversation until an authorized save installs it. This release introduces no new schema, save type, startup file, autonomous NPC process, cast-wide scan or automatic checkpoint.

Documentation and checks

The mechanics guide updates its agent-state diagram and includes a bare-approach example. All nine diagrams remain. Quick start, installation, upgrade guidance and architecture notes reflect the trigger.

Six optional B16 diagnostic cases cover ordering, undeclared intent, routine contact, NPC initiative, source distinctions, rules and persistence. They remain NOT RUN; structural tests and static review do not prove model compliance. The verification report separates observed engineering results from these pending behavioral observations.

RPG OS v0.9.0

RPG OS v0.9.0 — source access and evidence audits

This experimental release adds optional tools to help a GM find exact source passages and check a save against what actually happened in the session. It remains an unbound fresh install with Freeform and blank records. No Tellus module, campaign, character, artwork, raw conversation or private audit is included.

Added

• Exact source reads: section or original-line selections from the actual UTF-8 file, revision hashes, duplicate-heading detection, explicit truncation and optional delivery receipts. Section selection retains nested context and stops at the next same or shallower heading.
• Scoped lexical search: an optional SQLite FTS5 cache with original-file references, identities/aliases, revision checks and verified source fetching. Current state, world baselines, historical evidence, raw captures and rules have explicit scopes. An unusable cache leaves direct file access available.
• Untouched session capture: import real text or generic JSONL, preserve original bytes including OOC and rewinds, and declare source coverage. Missing exports remain missing; no transcript is reconstructed from summaries.
• Evidence-audit support: freeze selected prior/current sources and capture, generate a pending review template, then mechanically check a model- or human-authored report's hashes, quotations, coverage structure and repair gates. The helper does not perform semantic review or authorize repairs.
• Optional review after full saves: preserve needed prior records before replacement, complete the protected save and review the selected evidence. Report save completion separately from audit findings, unresolved questions and coverage. Existing CHECKPOINT behavior is unchanged.
• Concrete portrayal and audit guidance: compact NPC examples without mandatory mannerism quotas; checks for omitted promises, custody, resources and retired cues whose durable state still matters.

Preserved

Normal play uses the existing small startup packet and file authorities. Independent-agent baselines and the standing random fallback remain in place. No extra NPC database, background simulation, paid model API, embedding service or mandatory Python installation is introduced. The optional cache is disposable; current campaign records remain authoritative at their existing paths.

Compatible v0.7.3, v0.8.0, v0.8.1 and v8.1.1 campaigns retain their record formats. Use the protected upgrade guide in a separate copy. Installing this release does not enable automatic audits, change the agreement or regenerate NPC state.

Use it

Quick start · How it works, with diagrams · Source tools · Evidence audit · Verification

An audit compares prior canon, session evidence and the resulting save. It can detect disagreements that a save summary alone would miss, but its semantic reviewer remains fallible. Correct quotations do not guarantee correct interpretation. Complete chat export depends on the actual host; import integrity does not certify export completeness. Long-campaign and cross-model reliability still require playtesting.

The intended release number is v0.9.0. The source-access implementation was initially published under the incorrect label v9.0.0; this release corrects that numbering and its documentation links. It also fixes a Windows Python 3.12 compatibility issue that could mistake different path/handle timestamp representations for a changed source. Identity and content metadata still match across both views, while each view's timestamps remain checked before and after reading. No features or record formats are added by this correction.

The release workflow verifies the published v0.9.0 tag, commit and package assets before retiring only the mistaken v9.0.0 release and tag at their recorded commit. Commit history remains intact. Earlier v8.1.1 references are retained as historical release and schema identifiers. This remains an experimental release, with no claim of stable 1.0 readiness.

Design lessons from Claude Corpus Engine informed the optional source-access approach. Implementation and RPG OS integration use original code and wording under the repository's existing licenses.

RPG OS v0.8.1

RPG OS v0.8.1 — simple standing random fallback

This experimental patch adds a small fallback for eligible unauthored outcomes that have too little established basis for grounded judgment. A sparse NPC record no longer supplies a reason by itself to assume inaction or to invent a personality before resolving the relevant question.

The fresh-install package is RPG_OS_v0.8.1.zip, with SHA256SUMS. Extract it into its own folder and start with Quick start. The nine-diagram mechanics guide explains the procedure with contact, promise and secret-disclosure examples.

What changes in play

Established facts, applicable engine rules and other selected methods come first. Supported judgment remains sufficient. When an eligible outcome is still too sparse for grounded judgment, the selected standing oracle can resolve it directly: frame one concrete question and its interval, then obtain actual random input. The supplied convention is one d6, 1–3 No and 4–6 Yes. These even odds are a game convention, not measured probabilities of human behavior.

The result applies only to that question. No contact this week does not establish permanent dislike; disclosure to one recipient does not automatically inform every enemy. Further consequences follow actual circumstances and applicable rules. Repeated queries about the same resolved opportunity reuse its result. A relevant interruption returns the next reserved player decision and leaves later time uncommitted.

The GM retains the question, mapping, actual input and scoped result through existing owners at the next requested save when needed. A later audit reports an omitted resolution honestly; it cannot fabricate a past roll or supporting motive.

Select the method once

Say: “Use the simple d6 fallback as our standing method for eligible unresolved outcomes when grounded judgment is insufficient.” Include it in a new campaign's accepted setup proposal. For an existing campaign, adopt it prospectively through Recalibrate, retaining unrelated terms and already supplied acceptance. A standing selection needs no new permission for each later roll. If actual randomness is unavailable, obtain the player's roll or agree an available alternative.

Installing this patch alone does not alter an accepted diceless campaign, replace another selected oracle, reroll an established person or rewrite history. Missing required rules and inaccessible established facts remain retrieval or repair problems; deliberately unfixed facts retain their agreed triggers.

Compatibility and evidence

Current-save, agreement, people/system, archive and handover formats remain unchanged. The update adds no required NPC dossier, schema, startup read, cast-wide polling, recurring scheduler or automatic save. Unsaved play still depends on available conversation until an authorized save completes.

Use the protected v0.8 upgrade procedure for a compatible existing campaign. Preserve the complete old campaign and runtime, and apply the reviewed program files in a separate copy. Never extract blank INSTANCE or ARCHIVE templates over a played campaign.

The verification report separates previously observed engineering checks on the reviewed fallback candidate from final v0.8.1 checks. The nine optional B15 diagnostic variants and fresh-chat human replay remain NOT RUN. A successful structural check cannot establish reliable GM behavior, enjoyable pacing or lossless campaign continuity.

The public distribution remains an unbound fresh install with Freeform, generic module contracts and empty campaign/archive records. It includes no private campaign, character, transcript or artwork. Documentation/protocol material retains CC BY 4.0 licensing; code/configuration retains MIT.

RPG OS v0.8.0

RPG OS v0.8.0 — integrated agent state, experimental playtest release

This build integrates independent agent state throughout the campaign lifecycle and is ready for human campaign playtesting. It remains experimental: reliable NPC portrayal, enjoyable pacing and continuity across models must be assessed through actual play. No scripted behavioral trial schedule is a prerequisite for using it.

Download RPG_OS_v0.8.0.zip from this release and extract it into its own folder. Start with Quick start, read How RPG OS works, or use the upgrade guide for an existing campaign.

What changes in play

The GM establishes materially needed, genuinely unfixed agent facts through the accepted source or procedure before using them to resolve a response. It retains individual circumstances, aims, beliefs and relationships in their existing authorities and changes them for actual causes.

A courteous coworker can remain personally distant while developing professional trust. Relevant evidence can change a belief. An established goal and an actual opportunity can lead to initiative. Neither the player's desired outcome nor repeated equivalent attempts automatically create progress, resistance or a new random chance.

There is no compulsory trait vector, extra startup read, world-wide NPC tick, retrospective cast rewrite or automatic checkpoint.

Integrated lifecycle

• A compact consolidated core rule and cold OS/AGENT_STATE.md explain establishment, portrayal, causality, randomness and retention.
• Quick start, Guided and Detailed share one adaptive proposal and protected bind. Opening state uses the existing MODULE baseline; current people and systems retain their existing INSTANCE owners.
• Worldbuilding distinguishes capabilities, culture, individual facts, deliberately unfixed properties and missing sources. Source visuals govern their documented appearance scope, not invented personality.
• Random methods require defined applicable outcomes and actual input. Freeform judgment remains available and honestly labeled.
• Saving preserves established mixed relationships, qualified beliefs, relevant changes and sources. Fresh resumption and scene handover use those results instead of generating the person again.
• An interrupted setup retains its exact pending generation/result route through recovery. Restoring pre-bind files alone does not discard selected determinations or permit a silent reroll; completion or explicit setup cancellation/replacement resolves the pending bind.
• A v0.8 upgrade procedure preserves the old campaign and its runtime, applies reviewed generic changes in a separate copy and retains accepted state and history.
• An optional real-campaign playtest guide supports ordinary play and switching models at completed saves or through handover, with lightweight incident notes.
• The mechanics guide adds an agent-state lifecycle diagram; public setup, installation, commands and contribution guidance describe the implemented workflow.
• Structural, handover and packaging checks cover the new generic file inventory and relevant retained-record paths. Actual results are reported in Verification.
• Packaging preserves committed file bytes under Windows checkout preferences without changing Git settings. The public issue form accepts brief campaign incidents and fixes its version-placeholder indentation.

Retention and scope

Ordinary PLAY still writes no campaign files. New state remains in conversation until an authorized save completes. A requested CHECKPOINT preserves the complete accepted present and keeps the previous archive evidence boundary; it does not preserve missing transcript wording.

This release does not implement the proposed automatic protected-establishment checkpoint experiment, an additional live state layer, or an encrypted/private model memory. Source files remain frozen during scene handover; receiver-side developments must be captured in the permitted return. A genuinely lost determination is a retention gap, not permission to claim the original result was recovered.

The protocol preserves provider-neutral presentation wording and the campaign's accepted authorship and operator terms.

Compatibility and evidence

The v0.7.1–v0.7.3 current-save and agreement formats remain compatible. Existing people, module baselines and accepted relationships are not rerolled or retrospectively completed. See the upgrade guide; older formats retain their existing migration route.

Engineering checks remain part of delivery. Human campaign testing is the primary next evaluation of narrative quality and practical operation. Proposed behavioral cases and historical trials are not reported as new v0.8 results.

The distribution is an unbound fresh install containing Freeform and blank campaign/archive records. It includes no Tellus module, character, private campaign, transcript, personal policy or artwork. Documentation/protocol material retains CC BY 4.0 licensing; code/configuration retains MIT.

RPG OS v0.7.3

RPG OS v0.7.3 — live scene handover and illustrated mechanics

This is a clean, unbound fresh install with the Freeform engine. It contains no campaign module, character, session transcript, private GM preparation, or campaign artwork.

Continue a live scene with another GM

The optional scene-handover procedure freezes the current unresolved choice and separates the accessible conversation from established GM state. It preserves current mechanics, NPC knowledge and motives, hidden facts, elapsed time, pending events, and exact resume conditions where those are actually available.

The receiver explicitly accepts the transfer. The source stays paused while the receiving GM continues. A factual, non-graphic return supplies outcomes, evidence, and proposed state changes for a checked, recoverable import. This format governs the transfer record. No external service is called and no GM capability is assumed.

Manifests, export seals, hashes, identity checks, and receipts detect specified forms of drift, mismatch, and repeated import. Coverage labels expose incomplete conversation or unknown state; the mechanism does not claim to export a model's internal reasoning or reconstruct missing history.

Ordinary CHECKPOINT still preserves the present without advancing the previous archive evidence boundary. Handover is an additional live-state procedure, not a renamed save.

Understand the system

The new mechanics guide explains the operating loop, setup, player authorship, current and historical memory, private information, recovery, handover, and reference visuals. Seven Mermaid diagrams and a workspace tree show how those parts connect.

A prominent “How RPG OS works” card in the README provides front-page access. GitHub's built-in README, Contributing, and License tabs remain unchanged.

Runtime and release checks

• Provider-neutral runtime wording carries forward the current generic approach. Operator-accepted limits, player authorship, pause, correction, and retcon mechanisms remain available; no provider-specific depiction policy is bundled.
• The optional handover checker validates exported and returned packages. The main structural validator now identifies itself as VALIDATE-v3.0.2 and flags active handovers without claiming to perform the dedicated integrity check.
• All three Python regression suites run in CI.
• A new packaging helper exports committed Git content, enforces the public fresh-install inventory and unbound templates, validates the frozen tree, and checks ZIP contents and hashes.
• The release workflow publishes RPG_OS_v0.7.3.zip and SHA256SUMS. Local campaign folders, handovers, recovery files, build output, and untracked data are excluded.

See VERIFICATION for observed results and their limits.

Install or update

For a new campaign, extract the release into its own folder and follow QUICKSTART.

Valid v0.7.1/v0.7.2 campaigns need no current-state or agreement-format migration. Update generic runtime and tools in a protected copy, preserving campaign-owned modules, adapters, INSTANCE, and ARCHIVE. Do not install blank templates over a campaign, change pinned authorities during an active transfer, or discard recovery evidence. The optional handover grant can be accepted prospectively through the existing agreement procedure.

The kit remains experimental. Structural tests do not establish semantic losslessness, sustained campaign quality, or compatibility with every model host.