|
@@ -22,7 +22,7 @@ Open-source multi-agent system that autonomously writes, audits, and revises nov
|
|
|
|
|
|
|
|
## v0.4 Update
|
|
## v0.4 Update
|
|
|
|
|
|
|
|
-Spinoff writing + style cloning + post-write validator + audit-revise hardening + Writer two-phase split + multi-provider routing.
|
|
|
|
|
|
|
+Spinoff writing + style cloning + post-write validator + audit-revise hardening.
|
|
|
|
|
|
|
|
### Spinoff Writing
|
|
### Spinoff Writing
|
|
|
|
|
|
|
@@ -88,23 +88,6 @@ Real testing showed `rewrite` mode introduces 6x more AI markers than the origin
|
|
|
- Re-audit temperature locked to 0 (deterministic pass/fail gating)
|
|
- Re-audit temperature locked to 0 (deterministic pass/fail gating)
|
|
|
- `polish` mode boundaries strengthened (no adding paragraphs, renaming entities, or changing causality)
|
|
- `polish` mode boundaries strengthened (no adding paragraphs, renaming entities, or changing causality)
|
|
|
|
|
|
|
|
-### Writer Two-Phase Split
|
|
|
|
|
-
|
|
|
|
|
-Writer LLM calls are split into two phases: creative (temp 0.7, 8192 tokens) outputs prose only, settlement (temp 0.3, 8192 tokens) updates truth files only. Higher temperature for creativity, lower for factual accuracy. Chapter context compressed from last 3 chapters to last 1 + historical summaries.
|
|
|
|
|
-
|
|
|
|
|
-### Multi-Provider Routing
|
|
|
|
|
-
|
|
|
|
|
-Different agents can use different LLM providers (different baseUrl, API key, streaming settings). Use expensive models for Writer quality, cheap models for Auditor cost savings:
|
|
|
|
|
-
|
|
|
|
|
-```bash
|
|
|
|
|
-inkos config set-model auditor gemini-2.5-flash \
|
|
|
|
|
- --base-url https://www.packyapi.com/v1 \
|
|
|
|
|
- --api-key-env PACKYAPI_KEY \
|
|
|
|
|
- --no-stream
|
|
|
|
|
-
|
|
|
|
|
-inkos config show-models # View routing
|
|
|
|
|
-```
|
|
|
|
|
-
|
|
|
|
|
### Other v0.4 Changes
|
|
### Other v0.4 Changes
|
|
|
|
|
|
|
|
- Audit dimensions expanded from 26 to 33 (+4 spinoff dims + dim 27 sensitive words + dim 32 reader expectation management + dim 33 outline adherence detection)
|
|
- Audit dimensions expanded from 26 to 33 (+4 spinoff dims + dim 27 sensitive words + dim 32 reader expectation management + dim 33 outline adherence detection)
|
|
@@ -113,9 +96,6 @@ inkos config show-models # View routing
|
|
|
- Scheduler rewrite: AI-paced (15min cycles), parallel book processing, immediate retry, daily cap
|
|
- Scheduler rewrite: AI-paced (15min cycles), parallel book processing, immediate retry, daily cap
|
|
|
- New `spot-fix` revise mode (targeted repair)
|
|
- New `spot-fix` revise mode (targeted repair)
|
|
|
- `additionalAuditDimensions` in `book_rules.md` now supports name-string matching
|
|
- `additionalAuditDimensions` in `book_rules.md` now supports name-string matching
|
|
|
-- Non-streaming LLM support (per-provider `stream: false`)
|
|
|
|
|
-- Error diagnostics: no longer swallows all exceptions as "inkos.json not found"
|
|
|
|
|
-- CLI version dynamically read from package.json
|
|
|
|
|
|
|
|
|
|
---
|
|
---
|
|
|
|
|
|
|
@@ -412,7 +392,7 @@ inkos up # Daemon mode
|
|
|
| `inkos config set-global` | Set global LLM config (~/.inkos/.env) |
|
|
| `inkos config set-global` | Set global LLM config (~/.inkos/.env) |
|
|
|
| `inkos config show-global` | Show global config |
|
|
| `inkos config show-global` | Show global config |
|
|
|
| `inkos config set/show` | View/update project config |
|
|
| `inkos config set/show` | View/update project config |
|
|
|
-| `inkos config set-model <agent> <model>` | Set model override for a specific agent (`--base-url`/`--provider`/`--api-key-env`/`--no-stream`) |
|
|
|
|
|
|
|
+| `inkos config set-model <agent> <model>` | Set model override for a specific agent |
|
|
|
| `inkos config remove-model <agent>` | Remove agent model override (fall back to default) |
|
|
| `inkos config remove-model <agent>` | Remove agent model override (fall back to default) |
|
|
|
| `inkos config show-models` | Show current model routing |
|
|
| `inkos config show-models` | Show current model routing |
|
|
|
| `inkos doctor` | Diagnose setup issues (includes API connectivity test) |
|
|
| `inkos doctor` | Diagnose setup issues (includes API connectivity test) |
|
|
@@ -494,7 +474,7 @@ TypeScript monorepo managed with pnpm workspaces.
|
|
|
- [x] Style cloning (statistical fingerprint + LLM style guide + Writer injection)
|
|
- [x] Style cloning (statistical fingerprint + LLM style guide + Writer injection)
|
|
|
- [x] Post-write validator (11 hard rules + auto spot-fix)
|
|
- [x] Post-write validator (11 hard rules + auto spot-fix)
|
|
|
- [x] Audit-revise loop hardening (AI marker guard + temperature lock)
|
|
- [x] Audit-revise loop hardening (AI marker guard + temperature lock)
|
|
|
-- [x] Multi-LLM provider (OpenAI + Anthropic + compatible endpoints + per-agent provider routing)
|
|
|
|
|
|
|
+- [x] Multi-LLM provider (OpenAI + Anthropic + compatible endpoints)
|
|
|
- [x] AIGC detection + anti-detect rewrite pipeline
|
|
- [x] AIGC detection + anti-detect rewrite pipeline
|
|
|
- [x] Webhook notifications + smart scheduler (quality gates)
|
|
- [x] Webhook notifications + smart scheduler (quality gates)
|
|
|
- [x] Cross-chapter coherence (chapter summaries + subplot/emotion/character matrices)
|
|
- [x] Cross-chapter coherence (chapter summaries + subplot/emotion/character matrices)
|