Provides tools for inspecting Git diffs, unstaged changes, and commit history via the silvirica_git tool.
Provides an Obsidian-style markdown memory vault with wikilinks for storing architectural decisions, bug fixes, and failure signals, accessible via the silvirica_memory and silvirica_recall tools.
Integrates with Ollama as a local LLM provider, allowing model routing and optimized context delivery to Ollama models.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Silvirica where is the authenticate function defined and who calls it?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Zero-Model FastGate • >99% Token Reduction • Multi-Tier L1–L7 Cache • Obsidian Memory Vault • 142 Skills • Multi-IDE MCP
Silvirica AI is NOT another foundation model or conversational chatbot.
Silvirica AI is an AI Intelligence Optimization Layer that runs locally between your IDE / Coding Assistant / Agent and the LLM / AI Foundation Model.
Instead of naively feeding thousands of lines of raw code to expensive models, Silvirica:
FastGate Zero-Model Resolver: Instantly answers deterministic repository questions (symbol definitions, paths, dependency trees, git status) in <40ms with 0 tokens ($0.00).
AST & Knowledge Graph Compiler: Traverses multi-hop symbol relationships across Python, JS, TS, PHP, Go, Rust, and C++ to extract only the surgical AST signatures required.
Multi-Tier Semantic Cache (L1–L7): Caches deterministic symbol indexes, AST trees, file hashes, and query embeddings locally.
Obsidian-Style Markdown Memory Vault: Automatically preserves architectural decisions (ADRs), past bug fixes, and failure signals in .silvirica/memory/ across chat sessions.
Defensive Security & Shannon Entropy Redaction: Intercepts secrets, tokens, private keys, and prompt injection attacks before requests leave your machine.
Progressive 3-Tier Skill System: Selectively activates instructions across 142 domain skills without bloating token budgets.
┌─────────────────────────────────────────────────────────────────────────────┐
│ DEVELOPER / IDE / CODING AGENT │
│ (Cursor, Antigravity, Claude Code, VS Code, Cline, Zed) │
└──────────────────────────────────────┬──────────────────────────────────────┘
│
▼ (MCP stdio / JSON-RPC / REST)
┌─────────────────────────────────────────────────────────────────────────────┐
│ SILVIRICA OPTIMIZATION RUNTIME │
│ │
│ ┌───────────────────────┐ ┌─────────────────────────┐ ┌───────────────┐ │
│ │ FastGate & Zero-Model │ │ Multi-Language AST Index │ │ Smart Model │ │
│ │ (0 Tokens, <40ms) │ │ (Tree-Sitter & PyAST) │ │ Router (Emp.) │ │
│ └──────────┬────────────┘ └────────────┬────────────┘ └───────┬───────┘ │
│ │ │ │ │
│ ┌──────────▼────────────┐ ┌────────────▼────────────┐ ┌───────▼───────┐ │
│ │ SQLite Knowledge Graph│ │ Obsidian Memory Vault │ │ 142 Skills │ │
│ │ (Callers, Dependencies│ │ (12 Curated Categories) │ │ (Level 0/1/2) │ │
│ └──────────┬────────────┘ └────────────┬────────────┘ └───────┬───────┘ │
│ │ │ │ │
│ ┌──────────▼────────────────────────────▼───────────────────────▼───────┐ │
│ │ Context Compiler + Shannon Entropy Redactor + Adaptive Token Budget │ │
│ └───────────────────────────────────────┬───────────────────────────────┘ │
└──────────────────────────────────────────┼──────────────────────────────────┘
│ (Surgical Context: ~200 tokens)
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ LLM / FOUNDATION MODELS │
│ (Claude 3.7 Sonnet • GPT-4.5 • o3-mini • Gemini 2.0 • Ollama) │
└─────────────────────────────────────────────────────────────────────────────┘Related MCP server: codebase-rag
Feature Dimension | Without Silvirica AI (Naive Agent) | With Silvirica AI Runtime |
Token Consumption | 20,000–80,000 tokens / prompt (raw repo dump) | 50–400 tokens (Surgical AST & signature compiler) |
Deterministic Queries | Calls expensive LLM (takes 4–8s, costs $0.05–$0.20) | Zero-Model FastGate (<40ms, 0 tokens, $0.00) |
Project Memory | Lost immediately when chat context resets | Persistent Obsidian-style Markdown Vault & ADRs |
Symbol & Graph Navigation | Blind text search / Hallucinated imports | Multi-hop SQLite Knowledge Graph & Callers |
Skill Injection Overhead | 5,000+ tokens of static instructions every turn | Progressive 3-Tier Loading (Level 0 / 1 / 2) |
Defensive Security | API keys, passwords, and | Shannon Entropy & Regex local sanitization |
Model Cost Routing | 100% routed to flagship Tier-1 models ($$$) | Empirical Outcome Routing (Quick, Coder, Architect, Deep) |
Silvirica AI provides hardened, isolated installers for Windows (PowerShell) and POSIX systems (Linux & macOS), as well as direct Python package installation.
Stable Release vs. Development Channel:
Stable Releases (-Channel release / --channel release, default) use pinned release tags (e.g. v0.1.0) with optional SHA-256 checksum verification. The installer will never silently fall back to main.
Development Channel (-Channel main / --channel main) explicitly tracks the latest unreleased development branch.
Stable Release Installation (Recommended):
# Download installer
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/vinzz-yy/silvirica-ai/main/install.ps1" -OutFile "install.ps1"
# Verify & run installer with pinned release
powershell -ExecutionPolicy Bypass -File .\install.ps1 -Version "0.1.0"With SHA-256 Checksum Verification:
powershell -ExecutionPolicy Bypass -File .\install.ps1 -Version "0.1.0" -ExpectedSha256 "<64-hex-sha256-hash>"Development Channel (Latest main):
powershell -ExecutionPolicy Bypass -File .\install.ps1 -Channel mainDry-Run Verification:
powershell -ExecutionPolicy Bypass -File .\install.ps1 -DryRunStable Release Installation:
curl -fsSL https://raw.githubusercontent.com/vinzz-yy/silvirica-ai/main/install.sh -o install.sh
bash install.sh --version 0.1.0With SHA-256 Checksum Verification:
bash install.sh --version 0.1.0 --expected-sha256 "<64-hex-sha256-hash>"Development Channel (Latest main):
bash install.sh --channel maingit clone https://github.com/vinzz-yy/silvirica-ai.git
cd silvirica-ai
python -m venv venv
# On Windows: .\venv\Scripts\activate
# On Linux/macOS: source venv/bin/activate
pip install -e ".[dev]"Windows: Installed shims are located at %LOCALAPPDATA%\silvirica\bin. The installer adds this to User PATH non-destructively.
Linux/macOS: Symlinked to ~/.local/bin/silvirica. Ensure ~/.local/bin is in your PATH (export PATH="$HOME/.local/bin:$PATH").
Uninstallation:
Windows: Remove-Item -Recurse -Force "$env:LOCALAPPDATA\silvirica"
Linux/macOS: rm -rf ~/.local/share/silvirica ~/.local/bin/silvirica
Silvirica AI is designed with truthful provider reporting and an offline-first Local Deterministic Fallback Mode.
When no external provider API key is configured, Silvirica operates 100% locally:
Exact symbol searches, file listings, and call graphs are resolved via AST & SQLite Knowledge Graph in <40ms with 0 tokens.
Queries requiring reasoning provide relevant local context and symbol definitions without faking model responses.
explain-last accurately reports Zero-Model Hit: YES, Provider Called: NO, Actual Model: NONE.
To enable cloud reasoning tiers (e.g. OpenAI, Anthropic, or local Ollama endpoints):
# OpenAI or compatible endpoints
export OPENAI_API_KEY="sk-..."
export OPENAI_BASE_URL="https://api.openai.com/v1" # or http://localhost:11434/v1 for Ollama
# Anthropic endpoints
export ANTHROPIC_API_KEY="sk-ant-..."Silvirica AI natively implements the Model Context Protocol (MCP). Connect Silvirica to any IDE or agent in seconds:
mcp_config.json / claude_desktop_config.json){
"mcpServers": {
"silvirica": {
"command": "silvirica",
"args": ["mcp"]
}
}
}MCP Tool | Purpose | Typical Latency | Risk Rating |
| Auto-detects framework, languages, package managers, and architecture. |
|
|
| Hybrid structural and lexical code search with relevance scoring. |
|
|
| Surgical symbol signature and docstring extractor (Zero-Token Path). |
|
|
| Multi-hop dependency and caller graph traversal via SQLite. |
|
|
| Obsidian markdown memory vault reader with wikilinks. |
|
|
| Fast ADR, failure signal, and architectural precedent retrieval. |
|
|
| Progressive skill instructions loader (Level 1 Synopsis / Level 2 Deep). |
|
|
| Defensive vulnerability, permissions, and credential scanner. |
|
|
| Smart Context Compiler (budget allocation + entropy redaction). |
|
|
| JEV & empirical model selector and reasoning budget calculator. |
|
|
| Blast radius and multi-hop symbol change impact analyzer. |
|
|
| Git diff, unstaged change inspector, and commit history. |
|
|
| Real-time observatory telemetry, cache efficiency, and token savings. |
|
|
| Silvirica Doctor 18-point runtime diagnostic matrix. |
|
|
Silvirica AI features an enterprise-grade catalog of 142 progressive skills organized into 8 core engineering domains.
Unlike traditional assistants that stuff tens of thousands of instruction tokens into every prompt, Silvirica uses a Progressive 3-Tier Activation Pipeline:
Level 0 (Zero-Token Match): Triggers evaluated locally via AST and regex in <2ms with 0 tokens.
Level 1 (Surgical Synopsis): Injects ~30 tokens of high-signal rules when a skill is active.
Level 2 (Deep Execution Protocol): Full workflow instructions loaded on-demand via the silvirica_skill MCP tool.
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Summarize hermes-achievements badges, tiers, recent unlocks, and progress from local plugin artifacts |
|
| Coordinate multiple Hermes profiles or agents with task, handoff, heartbeat, blocker, and completion states |
|
| Write or update what an agent cannot derive from the code, inside a marked region, with every command verified o... |
|
| Consulting an external advisor when configured |
|
| Design recurring Hermes operations with schedule, delivery, silence policy, context chain, and prepared-vs-obser... |
|
| Policy overlay for browser tasks - add auth, confirmation, and observed-trace gates after preferring the native ... |
|
| Connect and operate Hermes as a native Buzz community agent, deliver local media with verified relay receipts, o... |
|
| Ending active workflow state cleanly |
|
| Turning one Silvirica capability family on or off so an install can be tailored instead of taken whole |
|
| Create a repo map, reading path, glossary, risk map, and first-task runway for unfamiliar codebases |
|
|
|
|
| Roadmap, PM, technical tradeoffs, risk, delivery, release, and follow-up operating cadence |
|
| Normalize Discord, Slack, Telegram, and other gateway sessions into origin, thread, delivery, silent, attachment... |
|
| Plan the resolution, the bisect, or the rewrite, name what is already pushed first, and force-push only with `--... |
|
| Secrets, outbound sends, blast radius; can only add a hold |
|
| Typed yes/no, pick-one, or scored questions to Jev with your own key; returns probabilities, never prose |
|
| Does the gathered evidence support the completion claim? It can only object |
|
| Auth, tests, migration risk, severity; flags only, never approves a merge |
|
| Answer an Silvirica route question about which workflow fits, recorded without re-routing |
|
| Turn an observed onboarding, activation, retention, re-engagement, referral, or monetization problem into one co... |
|
| Policy overlay for live lookups - add provider, freshness, units, and source-quality gates after preferring nati... |
|
| Prepare a build handoff for an LLM-powered feature with a pinned provider boundary, schema-first outputs, versio... |
|
| Reason over the imperative task, consult the live workflow catalog, and select or chain the right workflow(s) |
|
| Decide first whether prompting or retrieval already closes the gap, choose the method from the data you have, an... |
|
| Diagnose role-slot model configuration, guide provider connection, and apply changes only after diff approval |
|
| Meeting minutes, scrum/sprint records, retros, decisions, and follow-up history |
|
| Check version currency and parallel-tool capability status, then apply an update only after diff approval |
|
| Structured planning before execution |
|
| Test whether a customer problem, segment, and business hypothesis deserve product investment, ending in kill, pi... |
|
| Prompt import readiness - review and normalize external CLI-agent prompt files before offering slash-command can... |
|
| Refactor planning - turn a decided boundary-changing refactor into a phased plan - reconnaissance, contracts-fir... |
|
| Postmortems, SLOs, error budgets, incident follow-ups, and service reliability evidence |
|
| Router guidance for using silvirica-ai workflow skills inside Hermes Agent |
|
| Showing which coding units are running right now, on which runtime and model, with observed tokens and elapsed time |
|
| Managing local skills |
|
| Prepare a metadata-only Silvirica skill portfolio dashboard with stale surfaces, observed failure signals, pendi... |
|
| Prepare a metadata-only search-before-creation report for local, marketplace, GitHub, and web skill candidates w... |
|
| Turn a support case into a clear customer reply, severity path, and owned next step |
|
| Continue or finish the accepted work from conversation context, preserve rejected ideas, and report evidence-bou... |
|
| Diagnose scraper and auxiliary extract-model configuration, guide account setup, and apply each change as its ow... |
|
| Look up, capture, correct, and align the words a repository uses before planning or handoff |
|
| One-question-at-a-time clarification |
|
| Agentic interviewer -> planner -> researcher -> builder -> reviewer cycles until a real gate |
|
| Prepares the handoff for the coding agent you already chose, composing its prompt from that agent's own installe... |
|
| Consensus planning with review gates |
|
| Split it into disjoint parallel lanes with per-lane acceptance criteria, verification commands, and owners; prev... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Scope publish-ready writing, rewriting, summarization, translation, release-note, newsletter, customer-copy, soc... |
|
| Product identity, public capability catalog, model routing, local state, and long-term memory |
|
| Turn finance and accounting inputs into a decision-ready variance, cash, and close-risk brief |
|
| Turn repeated project or cross-project lessons into atomic, confidence-scored instinct candidates with scoped pr... |
|
| Select and confirm an immediate learning target, research credible sources, and prepare an application-first bri... |
|
| Read a very large PDF, contract, manual, or report through Hermes in page-anchored ranges with a coverage ledger |
|
| Agenda, prompts, decisions, and record template |
|
| Capture one bounded durable project or product memory candidate through explicit remember, refuse, or defer revi... |
|
| Inspect USER.md and MEMORY.md claims and prepare a native write diff without invoking, applying, or observing a ... |
|
| Morning brief SETUP (one-time) - connects mail and calendar MCP with read-and-draft-only scope and diff approval... |
|
| Explain a supplied paper or paper/PDF at a selected level while preserving full section coverage and source evid... |
|
| Turn product evidence into a decision-ready PRD, prioritization frame, and roadmap brief |
|
| Business research brief - turns a market, competitor, pricing, or customer question into a structured evidence-v... |
|
| Research operations department - coordinate Scout, Analyst, and Briefer work with source-inbox and status bounda... |
|
| Extract repeated principles from skills, prompts, traces, reviews, and failures into reviewed rule candidates wi... |
|
| Turn an account or market opportunity into a focused discovery, qualification, and next-step brief |
|
| Web lookup lane - settle a current-facts question in one cited retrieval round with retrieval dates and source-q... |
|
| Wiki construction blueprints and retained knowledge capture with destination-aware external knowledge connection... |
|
| Classify and review self-improvement store routes as an auxiliary review lane before durable writes, then record... |
|
| Study open-source reference implementations with pinned refs, gather live web evidence with citation discipline,... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| External app actions - email, Slack, Discord, Notion, Linear, Jira, CRM, and similar providers, scoped with auth... |
|
| Prepare server, API, and data-layer contracts — auth boundary, error paths, response shape, and schema/migration... |
|
| Turn a repository into one readable, interface-level PlantUML architecture picture - packages or modules, the pu... |
|
| Refresh local code intelligence, summarize repo structure, and prepare task-scoped codegraph handoff context wit... |
|
| Scope supplied data with provenance, causal-claim, and hallucination guards |
|
| Make every rerun idempotent, bound every replay, and gate each load on observed checks |
|
| Policy overlay for local file tasks - add path scoping and destructive-action gates after preferring native file... |
|
| Plan it with the checks that prove it safe, and never call a migration ready without its lock behaviour and roll... |
|
| Prepare Rust changes with ownership, error, and API discipline, and escalate any unsafe, FFI, or lock-free chang... |
|
| Pipeline-review, forecast-review, deal-review |
|
| Source candidate inventory - prepare typed source candidates and acquisition status before downstream work; use ... |
|
| Policy overlay for terminal commands - add cwd, environment, safety, and result-evidence gates after preferring ... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Prepare WCAG, keyboard, focus, screen-reader, target-size, and reflow evidence gates for UI surfaces |
|
| Turn a system's components and data flows into assets, trust boundaries, attack scenarios, controls, and the sec... |
|
| Find swallowed errors, unsafe fallbacks, hidden UI/runtime failures, and missing propagation before they become ... |
|
| Define the population and the sample, name the evidence that proves each item, re-perform the control, and grade... |
|
| Surface contract and compliance risks, questions, and escalation points before a legal decision or action |
|
| Evaluate release, deploy, security, observability, rollback, docs, and support readiness without claiming produc... |
|
| Triage reachability and severity, contain in order, and never close a leaked secret before its rotation is observed |
|
| Review prompt, tool, secret, dependency, destructive-action, and explicit local plugin risks before agent or cod... |
|
| Line citations, rank fixes and quick wins - and reconcile RESOLVED/NEW/CARRIED against the previous ledger on rerun |
|
| Map repository, skill, prompt, plugin, MCP, hook, config, and runtime surfaces before strengthening or operating... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Plan compact context, token/cost budgets, summarization checkpoints, and overflow recovery before long agent work |
|
| When a model family ships a new generation or changes its serving contract, walk the recognition, research, cali... |
|
| Report supplied local run efficiency while provider and host data stay unobserved |
|
| Find where a system is actually slow, leaking, or expensive across runtime, memory, token cost, storage, renderi... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Prepare native Apple UI or Apple marketing product-visual direction, review, and improvement briefs with evidenc... |
|
| Classify build, typecheck, lint, test, CI, and DCO failures into minimal safe fix handoffs |
|
| Turn a learning goal into a teachable curriculum, assessment plan, and learner-ready sequence |
|
| Prepare a bounded design direction, existing-lane composition, and executor-neutral handoff |
|
| Enforce superior content, design, layout, publishing, and visual QA gates |
|
| Prepare design-system-driven web and terminal (TUI) UI creation, redesign, polish, accessibility, performance, a... |
|
| Behavior-preserving refactor of UI code - preview the full change plan first, apply as a second explicit step, a... |
|
| Image prompt cards - turn meetings, reports, PRs, issues, research, and releases into domain-aware image prompt ... |
|
| Make a product or content release locale-ready with terminology, cultural-fit, and quality-review guidance |
|
| User-sent media - audio, video, YouTube links, screenshots, receipts, OCR, meeting recordings, transcripts, time... |
|
| Prepare observed-only rendered QA gates for web, frontend, image, document, and TUI surfaces |
|
| Terse voice and mobile-style requests - turn short spoken-style asks into clarify, plan, status, handoff, or con... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Capture a stuck, looping, drifting, or repeatedly failing agent run, diagnose the likely failure pattern, and pr... |
|
| Reproduce it first, form competing hypotheses, discriminate them with the cheapest observation, and only then fi... |
|
| Cluster customer signals and choose the next workflow |
|
| Retry, fix a dependency, ask for access, or change approach on a failing run |
|
| Command an incident that is still open -- severity as declared live state, commander and roles, an append-only t... |
|
| Prepare hypothesis-driven debugging of native binaries and instruct the executor to drive a DAP debugger instead... |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Independent perspectives attack a proposal, then distill into a bundle a separate planner consumes |
|
| Compare executor or agent choices on reproducible tasks using quality, cost, time, tool, and evidence metrics |
|
| Delete AI-generated slop, dead code, and duplication while observable behavior stays identical |
|
| Score a web surface against published design-award judging axes and name the binding constraint |
|
| Bug-first review with evidence |
|
| Define and record build, lint, typecheck, test, security, docs, generated-output, and CI evidence before complet... |
|
| Adversarial QA and fix loops |
|
Skill Identifier | Purpose & Capability | Auto-Activation Triggers |
| Help managers inspect AI-agent progress, blockers, quality gates, and throughput levers |
|
| Tradeoffs, a recommendation, and a decision note you can act on |
|
| Resolve one uncertain interaction, API, performance, or integration choice with a disposable, isolated experimen... |
|
| Recall scoped reviewed rejected decisions without elevating them to approved memory |
|
| Track PPT, PDF, XLSX, DOCX, HWP, Markdown, and attachments through prepared, generated, QA, approved, and attach... |
|
| Release checklist, deploy decision, health signals, rollback gate, and post-deploy status |
|
| Diagnosing silvirica-ai installation health |
|
| Executor runtime readiness - compare Codex, Claude Code, Hermes coding, and oh-my runtimes by tools and handoff ... |
|
| External connector readiness - assess whether a named plugin, connector, API, data provider, or multimodal route... |
|
| Route PR, issue, CI, and review webhook events into triage, review, or fix handoff cards |
|
| Turn a public chat report into a confirmed, verified issue package |
|
| Normalize Codex, Claude Code, Hermes, OpenCode, Cursor, MCP host, worktree, and wrapper session metadata into on... |
|
| Read the drift, the blast radius and the cost delta from the saved plan, then stage the apply behind a health ga... |
|
| Shape an app idea into decisions, delivery handoff, verification, release, and monitoring status |
|
| Choose the serving engine and quantization from decision tables, prepare deployment as an idempotent runbook wit... |
|
| Decks, PDFs, spreadsheets, documents, HWP, Markdown, and binary export handoffs |
|
| Prepare each gate the store enforces, and plan the halt and the next build before the rollout starts, because a ... |
|
| Prepare an operations command-board for wrapper-safe token, cost, latency, run history, queue, failure-mode, ext... |
|
| Status, risks, blockers, priorities, and follow-ups |
|
| Turn hiring and people context into a fair, structured recruiting or people-operations brief |
|
| Gate robots, 3D printers, IoT relays, sensors, and lab hardware before trials; use external-connector-readiness ... |
|
| Prepare provider-profile metadata without reading secrets or calling providers |
|
| Decide what goes in, the version, the rollout stages, and a rollback with its trigger and exact command before i... |
|
| Weekly/monthly reports, executive briefs, PPT-ready outlines, and upload packages |
|
| Inventory which MCP servers, CLIs, APIs, credentials, and connectors a workflow needs; use external-connector-re... |
|
Command | Description | Example |
| Initializes |
|
| Runs 18-point comprehensive runtime diagnostics (DBs, MCP, AST, Security, Cache). |
|
| Indexes repository symbols and builds knowledge graph. |
|
| Displays indexed symbols, graph nodes, memory notes, and cache stats. |
|
| Queries intelligence with Zero-Model resolver and Smart Context Compiler. |
|
| Queries multi-hop SQLite dependency and caller relationships. |
|
| Truthfully explains last routing decision, model tier, and retrieval counts. |
|
| Shows JEV capability health, circuit breaker state, and metrics. |
|
| Runs comprehensive diagnostics on JEV connectivity & latency. |
|
| Runs comparative benchmark: Without JEV vs. With JEV. |
|
| Scans workspace for exposed secrets, high-entropy tokens, and OWASP risks. |
|
| Lists all 142 progressive skills and audits overlapping triggers. |
|
| Inspects and searches Obsidian memory vault with wikilinks. |
|
| Inspects multi-tier L1-L7 hit rate statistics or purges caches. |
|
| Audits design system tokens, typography hierarchy, and CSS variables. |
|
| Launches real-time terminal Observatory telemetry HUD. |
|
| Runs verified benchmark harness with correctness gates. |
|
| Starts standard MCP stdio server for IDE integrations. |
|
| Launches background REST API daemon for extensions. |
|
Silvirica AI benchmarks compare real-world efficiency with strict Correctness Gates. A benchmark run only counts as an efficiency improvement if the output meets ground-truth task completion requirements (GATE: PASS).
==========================================================================================
SILVIRICA AI BENCHMARK SCOREBOARD
(BASELINE vs SURGICAL SILVIRICA PIPELINE)
==========================================================================================
SCENARIO | BASELINE (Tokens) | SILVIRICA CONTEXT | REDUCTION | GATE
------------------------------------------------------------------------------------------
Symbol Lookup (Where is AuthCon | 18,500 | 327 | 98.2% | PASS
File Path Discovery | 14,000 | 272 | 98.1% | PASS
Bug Investigation (Login Loop) | 34,000 | 290 | 99.1% | PASS
Feature Implementation (Rate Li | 42,000 | 282 | 99.3% | PASS
Large-Repository Dependency Deb | 68,000 | 285 | 99.6% | PASS
Defensive Security & IDOR Revie | 56,000 | 1,105 | 98.0% | PASS
Architecture Decision Memory An | 28,000 | 322 | 98.9% | PASS
Codebase Modular Refactoring | 48,000 | 1,135 | 97.6% | PASS
==========================================================================================
TOTAL BASELINE CONTEXT: 308,500 tokens
TOTAL SILVIRICA CONTEXT: 4,018 tokens
TOTAL CONTEXT REDUCED: 304,482 tokens (TOTAL TOKENS SAVED: 304,482)
OVERALL CONTEXT REDUCTION: 98.7% (OVERALL TOKEN REDUCTION: 98.7%)
CORRECTNESS GATE STATUS: ALL GATES PASSED (Verified)
==========================================================================================silvirica doctor reports "Local Deterministic Fallback Mode active"Cause: No external OPENAI_API_KEY or ANTHROPIC_API_KEY is configured in environment variables.
Solution: This is normal and intentional for offline use. Silvirica's Zero-Model FastGate and AST/Graph queries will function 100% locally with 0 tokens. To enable cloud LLM reasoning, export OPENAI_API_KEY or ANTHROPIC_API_KEY.
Run silvirica analyze --force to re-index all symbols and update the project state hash.
To clear all multi-tier caches manually: silvirica cache --clear.
Run with bypass: powershell -ExecutionPolicy Bypass -File .\install.ps1.
Silvirica does not require Administrator privileges and installs cleanly to %LOCALAPPDATA%\silvirica.
Silvirica AI is built with privacy-first engineering:
Local Execution: All symbol indexing, AST parsing, and knowledge graphs run 100% locally via SQLite.
Shannon Entropy Redaction: High-entropy strings (hex $\ge 3.6$, base64 $\ge 4.3$) and private keys are redacted before prompt compilation.
Strict Origin CORS: Local daemon enforces strict localhost origin checking and bearer token authorization.
Path Traversal Protection: All memory reads/writes are strictly sandboxed inside .silvirica/memory/.
System Security Directives: Context compiler injects explicit system directives declaring repository code as untrusted data to prevent indirect prompt injection.
Silvirica AI is open-source software licensed under the MIT License.
Built for engineers who demand maximum intelligence, ultra-low latency, and uncompromising security.
This server cannot be deployed
Search indexed code, trace dependencies, assess change impact, and recall repository memory.
Shared memory for coding agents. Stop re-explaining your codebase every session.
Code intelligence for coding agents: semantic, AST, graph, and full-text search. 279+ languages.
Give your AI agent a persistent map of your project's structure, dependencies, and bugs.