lynkr 9.1.9 → 9.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +211 -215
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -6,7 +6,6 @@
6
6
  [![Tests](https://img.shields.io/badge/tests-699%20passing-brightgreen)](https://github.com/Fast-Editor/Lynkr)
7
7
  [![License: Apache 2.0](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](LICENSE)
8
8
  [![Node.js](https://img.shields.io/badge/node-20%2B-green)](https://nodejs.org)
9
- [![Homebrew Tap](https://img.shields.io/badge/homebrew-lynkr-brightgreen.svg)](https://github.com/vishalveerareddy123/homebrew-lynkr)
10
9
  [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/vishalveerareddy123/Lynkr)
11
10
 
12
11
  <table>
@@ -20,289 +19,281 @@
20
19
 
21
20
  ---
22
21
 
23
- ## The Problem
22
+ ## Quick Start (2 Minutes)
24
23
 
25
- AI coding tools lock you into one provider. Claude Code requires Anthropic. Codex requires OpenAI. You can't use your company's Databricks endpoint, your local Ollama models, or your AWS Bedrock account — at least, not without Lynkr.
26
-
27
- **The real costs:**
28
- - Anthropic API at $15/MTok output adds up fast for daily coding
29
- - No way to use free local models (Ollama, llama.cpp) with Claude Code
30
- - Enterprise teams can't route through their own cloud infrastructure
31
- - Provider outages take your entire workflow down
32
-
33
- ## The Solution
34
-
35
- Lynkr is a self-hosted proxy that sits between your AI coding tools and any LLM provider. One environment variable change, and your tools work with any model.
36
-
37
- ```
38
- Claude Code / Cursor / Codex / Cline / Continue / Vercel AI SDK
39
- |
40
- Lynkr
41
- |
42
- Ollama | Bedrock | Databricks | OpenRouter | Azure | OpenAI | llama.cpp
43
- ```
24
+ ### 1. Install Lynkr
44
25
 
45
26
  ```bash
46
- # That's it. Three lines.
47
27
  npm install -g lynkr
48
- export ANTHROPIC_BASE_URL=http://localhost:8081
49
- lynkr start
50
28
  ```
51
29
 
52
- ---
30
+ ### 2. Choose Your Setup
53
31
 
54
- ## Quick Start
32
+ **Option A: Free & Local (Ollama)**
33
+ ```bash
34
+ # Install Ollama first: https://ollama.com
35
+ ollama pull qwen2.5-coder:latest
55
36
 
56
- ### Install
37
+ # Start Lynkr
38
+ lynkr start
39
+ ```
57
40
 
58
- **One-line install (recommended):**
41
+ Your `.env` file (auto-created on first run):
59
42
  ```bash
60
- curl -fsSL https://raw.githubusercontent.com/Fast-Editor/Lynkr/main/install.sh | bash
43
+ MODEL_PROVIDER=ollama
44
+ FALLBACK_ENABLED=false
45
+ OLLAMA_MODEL=qwen2.5-coder:latest
61
46
  ```
62
47
 
63
- **Or via npm:**
48
+ **Option B: Cloud Provider (OpenRouter)**
64
49
  ```bash
65
- npm install -g pino-pretty && npm install -g lynkr
50
+ # Get API key from https://openrouter.ai
51
+ lynkr start
66
52
  ```
67
53
 
68
- ### Pick a Provider
69
-
70
- **Free & Local (Ollama)**
54
+ Your `.env` file:
71
55
  ```bash
72
- export MODEL_PROVIDER=ollama
73
- export OLLAMA_MODEL=qwen2.5-coder:latest
74
- lynkr start
56
+ MODEL_PROVIDER=openrouter
57
+ OPENROUTER_API_KEY=sk-or-v1-your-key-here
58
+ FALLBACK_ENABLED=false
75
59
  ```
76
60
 
77
- **AWS Bedrock (100+ models)**
61
+ **Option C: Enterprise (AWS Bedrock)**
78
62
  ```bash
79
- export MODEL_PROVIDER=bedrock
80
- export AWS_BEDROCK_API_KEY=your-key
81
- export AWS_BEDROCK_MODEL_ID=anthropic.claude-3-5-sonnet-20241022-v2:0
82
63
  lynkr start
83
64
  ```
84
65
 
85
- **OpenRouter (cheapest cloud)**
66
+ Your `.env` file:
86
67
  ```bash
87
- export MODEL_PROVIDER=openrouter
88
- export OPENROUTER_API_KEY=sk-or-v1-your-key
89
- lynkr start
68
+ MODEL_PROVIDER=bedrock
69
+ AWS_BEDROCK_API_KEY=your-aws-key
70
+ AWS_BEDROCK_MODEL_ID=anthropic.claude-3-5-sonnet-20241022-v2:0
71
+ FALLBACK_ENABLED=false
90
72
  ```
91
73
 
92
- ### Connect Your Tool
74
+ ### 3. Connect Your Tool
93
75
 
94
76
  **Claude Code**
95
77
  ```bash
96
78
  export ANTHROPIC_BASE_URL=http://localhost:8081
97
79
  export ANTHROPIC_API_KEY=dummy
98
- claude "Your prompt here"
80
+ claude "write a hello world in python"
99
81
  ```
100
82
 
101
- **Codex CLI** — edit `~/.codex/config.toml`:
83
+ **Cursor IDE**
84
+ - Settings → Models → Override Base URL
85
+ - Set to: `http://localhost:8081/v1`
86
+ - API Key: `any-value`
87
+
88
+ **Codex CLI**
89
+
90
+ Edit `~/.codex/config.toml`:
102
91
  ```toml
103
92
  model_provider = "lynkr"
104
- model = "gpt-4o"
105
93
 
106
94
  [model_providers.lynkr]
107
- name = "Lynkr Proxy"
108
95
  base_url = "http://localhost:8081/v1"
109
96
  wire_api = "responses"
110
97
  ```
111
98
 
112
- **Cursor IDE**
113
- - Settings > Features > Models
114
- - Base URL: `http://localhost:8081/v1`
115
- - API Key: `sk-lynkr`
116
-
117
- **Vercel AI SDK**
118
- ```ts
119
- import { generateText } from "ai";
120
- import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
121
-
122
- const lynkr = createOpenAICompatible({
123
- baseURL: "http://localhost:8081/v1",
124
- name: "lynkr",
125
- apiKey: "sk-lynkr",
126
- });
127
-
128
- const { text } = await generateText({
129
- model: lynkr.chatModel("auto"),
130
- prompt: "Hello!",
131
- });
132
- ```
99
+ ✅ **Done!** Your AI tool now uses your chosen provider.
100
+
101
+ ---
102
+
103
+ ## Why Lynkr?
133
104
 
134
- **OpenClaw**
135
- ```json
136
- // openclaw.json
137
- {
138
- "models": {
139
- "providers": [{
140
- "name": "lynkr",
141
- "type": "openai-compatible",
142
- "base_url": "http://localhost:8081/v1",
143
- "api_key": "any-value",
144
- "models": ["auto"]
145
- }]
146
- }
147
- }
105
+ AI coding tools lock you into one provider. Lynkr breaks that lock.
106
+
107
+ ```
108
+ Claude Code / Cursor / Codex / Cline / Continue
109
+ ↓
110
+ Lynkr
111
+ ↓
112
+ Ollama | Bedrock | Azure | OpenRouter | OpenAI
148
113
  ```
149
- Set `OPENCLAW_MODE=true` in Lynkr's `.env` to show actual provider/model in responses.
150
114
 
151
- > Works with any OpenAI-compatible client: Cline, Continue.dev, OpenClaw, KiloCode, and more.
115
+ **What you get:**
116
+ - ✅ Use **free local models** (Ollama, llama.cpp) with Claude Code
117
+ - ✅ Route through **your company's infrastructure** (Databricks, Azure, Bedrock)
118
+ - ✅ Cut costs **60-80%** with smart token optimization
119
+ - ✅ **Zero code changes** - just change one environment variable
152
120
 
153
121
  ---
154
122
 
155
123
  ## Supported Providers
156
124
 
157
- | Provider | Type | Models | Cost |
158
- |----------|------|--------|------|
159
- | **Ollama** | Local | Unlimited (free, offline) | **Free** |
125
+ | Provider | Type | Example Models | Cost |
126
+ |----------|------|---------------|------|
127
+ | **Ollama** | Local | qwen2.5-coder, deepseek-coder, llama3 | **Free** |
160
128
  | **llama.cpp** | Local | Any GGUF model | **Free** |
161
129
  | **LM Studio** | Local | Local models with GUI | **Free** |
162
- | **MLX Server** | Local | Apple Silicon optimized | **Free** |
163
- | **AWS Bedrock** | Cloud | 100+ (Claude, Llama, Mistral, Titan) | $$ |
164
- | **OpenRouter** | Cloud | 100+ (GPT, Claude, Llama, Gemini) | $-$$ |
130
+ | **OpenRouter** | Cloud | GPT-4o, Claude 3.5, Llama 3, Gemini | $ |
131
+ | **AWS Bedrock** | Cloud | Claude, Llama, Mistral, Titan | $$ |
165
132
  | **Databricks** | Cloud | Claude Sonnet 4.5, Opus 4.6 | $$$ |
166
133
  | **Azure OpenAI** | Cloud | GPT-4o, o1, o3 | $$$ |
167
- | **Azure Anthropic** | Cloud | Claude models | $$$ |
168
- | **OpenAI** | Cloud | GPT-4o, o3, o4-mini | $$$ |
169
- | **Google Vertex** | Cloud | Gemini 2.5 Pro/Flash | $$$ |
170
- | **Moonshot AI** | Cloud | Kimi K2 Thinking/Turbo | $$ |
171
- | **Z.AI** | Cloud | GLM-4.7 | $$ |
172
- | **DeepSeek** | Cloud | DeepSeek Reasoner, R1 | $ |
134
+ | **Azure Anthropic** | Cloud | Claude Sonnet, Opus | $$$ |
135
+ | **OpenAI** | Cloud | GPT-4o, o3-mini | $$$ |
136
+ | **DeepSeek** | Cloud | DeepSeek R1, Reasoner | $ |
173
137
 
174
- 4 local providers for **100% offline, free** usage. 10+ cloud providers for scale.
138
+ **4 local providers** for 100% offline, free usage. **10+ cloud providers** for scale.
175
139
 
176
140
  ---
177
141
 
178
- ## Why Lynkr Over Alternatives
179
-
180
- | Feature | Lynkr | LiteLLM (42K stars) | OpenRouter | PortKey |
181
- |---------|-------|---------------------|------------|---------|
182
- | **Setup** | `npm install -g lynkr` | Python + Docker + Postgres | Account signup | Docker + config |
183
- | **Claude Code support** | Drop-in, native | Requires config | No CLI support | Requires config |
184
- | **Cursor support** | Drop-in, native | Partial | Via API key | Partial |
185
- | **Codex CLI support** | Drop-in, native | No | No | No |
186
- | **Built for coding tools** | Yes (purpose-built) | No (general gateway) | No (general API) | No (general gateway) |
187
- | **Local models** | Ollama, llama.cpp, LM Studio, MLX | Ollama only | No | No |
188
- | **Token optimization** | Built-in (60-80% savings) | No | No | Caching only |
189
- | **Complexity routing** | Auto-routes by task difficulty | Manual | Cost/latency only | Manual |
190
- | **Memory system** | Titans-inspired long-term memory | No | No | No |
191
- | **Self-hosted** | Yes (Node.js) | Yes (Python stack) | No (SaaS) | Yes (Docker) |
192
- | **Offline capable** | Yes | Yes | No | No |
193
- | **Transaction fees** | None | None (OSS) / Paid enterprise | 5.5% on credits | Free tier / Paid |
194
- | **Dependencies** | Node.js only | Python, Prisma, PostgreSQL | N/A | Docker, Python |
195
- | **Format conversion** | Anthropic <-> OpenAI (automatic) | Automatic | N/A | Automatic |
196
- | **Code intelligence** | Graphify (19-lang AST graph) | No | No | No |
197
- | **Routing telemetry** | Built-in (SQLite + REST API) | No | Dashboard | Dashboard |
198
- | **Admin hot-reload** | Yes (no restart) | Requires restart | N/A | Requires restart |
199
- | **License** | Apache 2.0 | MIT | Proprietary | MIT (gateway) |
200
-
201
- **Lynkr's edge:** Purpose-built for AI coding tools. Not a general LLM gateway — a proxy that understands Claude Code, Cursor, and Codex natively, with built-in token optimization, complexity-based routing, and a memory system designed for coding workflows. Installs in one command, runs on Node.js, zero infrastructure required.
142
+ ## Advanced: Tier Routing (Save Even More)
202
143
 
203
- ---
144
+ Route different request types to different models automatically:
204
145
 
205
- ## Cost Comparison
146
+ ```bash
147
+ # .env file
148
+ MODEL_PROVIDER=ollama
149
+ FALLBACK_ENABLED=false
206
150
 
207
- | Scenario | Direct Anthropic | Lynkr + Ollama | Lynkr + OpenRouter | Lynkr + Bedrock |
208
- |----------|-----------------|----------------|--------------------| --------------- |
209
- | Daily Claude Code usage | ~$10-30/day | **$0 (free)** | ~$2-8/day | ~$5-15/day |
210
- | Token optimization savings | — | — | 60-80% further | 60-80% further |
211
- | Monthly (heavy use) | $300-900 | **$0** | $60-240 | $150-450 |
151
+ # Use small/fast models for simple tasks
152
+ TIER_SIMPLE=ollama:qwen2.5:3b
212
153
 
213
- > With token optimization enabled, Lynkr's smart tool selection, prompt caching, and memory deduplication reduce token usage by 60-80% on top of provider savings.
154
+ # Use medium models for normal coding
155
+ TIER_MEDIUM=ollama:qwen2.5:7b
156
+
157
+ # Use powerful models for complex architecture
158
+ TIER_COMPLEX=ollama:deepseek-r1:14b
159
+ TIER_REASONING=ollama:deepseek-r1:14b
160
+
161
+ # Increase limits for long conversations
162
+ POLICY_MAX_STEPS=50
163
+ POLICY_MAX_TOOL_CALLS=100
164
+ ```
165
+
166
+ Lynkr analyzes each request and routes it to the appropriate tier. Simple questions use fast models. Complex refactoring uses powerful models.
167
+
168
+ **Result:** 70-90% of requests use cheaper/faster models. Only hard problems hit expensive models.
214
169
 
215
170
  ---
216
171
 
217
- ## What's Under the Hood
172
+ ## Configuration Examples
218
173
 
219
- Lynkr isn't just a passthrough proxy. It's an optimization layer.
174
+ ### Minimal Ollama Setup
175
+ ```bash
176
+ # .env
177
+ MODEL_PROVIDER=ollama
178
+ FALLBACK_ENABLED=false
179
+ OLLAMA_MODEL=qwen2.5-coder:latest
180
+ POLICY_MAX_STEPS=50
181
+ POLICY_MAX_TOOL_CALLS=100
182
+ ```
220
183
 
221
- ### Smart Routing (5-Phase)
222
- Routes requests to the right model based on 5-phase complexity analysis. Simple questions go to fast/cheap models. Complex architectural tasks go to powerful models. Includes Graphify structural analysis for code-aware routing.
184
+ ### Production Cloud Setup
185
+ ```bash
186
+ # .env
187
+ MODEL_PROVIDER=openrouter
188
+ OPENROUTER_API_KEY=sk-or-v1-your-key
189
+ FALLBACK_ENABLED=false
190
+
191
+ # Tier routing for cost optimization
192
+ TIER_SIMPLE=openrouter:google/gemini-flash-1.5
193
+ TIER_MEDIUM=openrouter:anthropic/claude-3.5-sonnet
194
+ TIER_COMPLEX=openrouter:anthropic/claude-opus-4
195
+ TIER_REASONING=openrouter:anthropic/claude-opus-4
196
+
197
+ # Increase limits
198
+ POLICY_MAX_STEPS=50
199
+ POLICY_MAX_TOOL_CALLS=100
200
+
201
+ # Optional: Enable caching for repeated prompts
202
+ PROMPT_CACHE_ENABLED=true
203
+ SEMANTIC_CACHE_ENABLED=true
204
+ ```
223
205
 
224
- - **Complexity scoring** — 15-dimension weighted scoring with agentic workflow detection
225
- - **Graphify integration** — AST-based knowledge graph detects god nodes, community cohesion, blast radius across 19 languages
226
- - **Routing telemetry** — every decision recorded with quality scoring (0-100) and latency tracking (P50/P95/P99)
206
+ ### Enterprise Databricks Setup
207
+ ```bash
208
+ # .env
209
+ MODEL_PROVIDER=databricks
210
+ DATABRICKS_API_BASE=https://your-workspace.cloud.databricks.com
211
+ DATABRICKS_API_KEY=your-token
212
+ FALLBACK_ENABLED=false
213
+
214
+ POLICY_MAX_STEPS=50
215
+ POLICY_MAX_TOOL_CALLS=100
216
+ ```
217
+
218
+ ---
227
219
 
228
- ### Token Optimization (8 Phases)
229
- - **MCP Code Mode** — replaces 100+ MCP tool schemas with 4 meta-tools (~96% reduction, lazy tool discovery)
230
- - **Smart tool selection** — only sends tools relevant to the current task (50-70% reduction)
231
- - **Prompt caching** — SHA-256 keyed LRU cache (30-45% reduction on repeated prompts)
232
- - **Memory deduplication** — eliminates repeated information across turns (20-30% reduction)
233
- - **Tool response truncation** — intelligent truncation of long outputs (15-25% reduction)
234
- - **Dynamic system prompts** — adapt complexity to request type (10-20% reduction)
235
- - **Distill compression** — structural similarity, delta rendering, smart dedup of repetitive tool outputs (20-40% reduction)
236
- - **Headroom sidecar** — optional ML-based compression: Smart Crusher, CCR, LLMLingua (47-92% reduction)
220
+ ## Common Issues & Fixes
237
221
 
238
- ### Enterprise Resilience
239
- - **Circuit breakers** — automatic failover with half-open probe recovery
240
- - **Admin hot-reload** — `POST /v1/admin/reload` reloads config + resets circuit breakers without restart
241
- - **Load shedding** — graceful degradation under high load
242
- - **Prometheus metrics** — full observability at `/metrics`
243
- - **Health checks** — K8s-ready endpoints at `/health`
244
- - **Performance timer** — per-request timing breakdown with `PERF_TIMER=true`
222
+ | Issue | Solution |
223
+ |-------|----------|
224
+ | **"Service temporarily overloaded"** | Ollama model too large for RAM. Use smaller model or increase `--max-old-space-size` |
225
+ | **"Route not found: HEAD /"** | Ignore - harmless health check from Claude Code |
226
+ | **"Hallucinated tool calls"** | Normal - Lynkr automatically filters invalid tools |
227
+ | **"Safe Command DSL blocked"** | Add `POLICY_SAFE_COMMANDS_ENABLED=false` to `.env` |
228
+ | **Slow first request (20+ sec)** | Ollama loading model into memory. Add `OLLAMA_KEEP_ALIVE=30m` in Ollama config |
229
+ | **No response after N turns** | Increase limits: `POLICY_MAX_STEPS=50` and `POLICY_MAX_TOOL_CALLS=100` |
245
230
 
246
- ### Memory System
247
- Titans-inspired long-term memory with surprise-based filtering. The system remembers important context across sessions and forgets noise — reducing token waste from repeated context.
231
+ ---
248
232
 
249
- ### Semantic Cache
250
- Cache responses for semantically similar prompts. Hit rate depends on your workflow, but repeat questions (common in coding) get instant responses.
233
+ ## Advanced Features
251
234
 
235
+ ### Token Optimization (60-80% savings)
252
236
  ```bash
237
+ # Enable all optimizations
238
+ PROMPT_CACHE_ENABLED=true
253
239
  SEMANTIC_CACHE_ENABLED=true
254
- SEMANTIC_CACHE_THRESHOLD=0.95
240
+ TOOL_INJECTION_ENABLED=false
241
+ CODE_MODE_ENABLED=true
255
242
  ```
256
243
 
257
- ### MCP Integration + Code Mode
258
- Automatic Model Context Protocol server discovery and orchestration. Your MCP tools work through Lynkr without configuration.
259
-
260
- **MCP Code Mode** — Token optimization for heavy MCP setups:
261
- - Replaces 100+ individual MCP tool schemas with 4 meta-tools
262
- - Reduces tool catalog from ~17,500 tokens to ~700 tokens (**96% reduction**)
263
- - Enables lazy tool discovery: model queries `mcp_list_tools`, then `mcp_tool_info`, then `mcp_execute`
264
- - Best for: 50+ MCP tools, long conversations, context-constrained setups
265
- - Trade-off: 3 sequential calls instead of 1 (adds ~2-3s latency)
244
+ ### Memory System (Titans-inspired)
245
+ ```bash
246
+ MEMORY_ENABLED=true
247
+ MEMORY_TTL=3600000 # 1 hour
248
+ ```
266
249
 
250
+ ### Load Shedding & Resilience
267
251
  ```bash
268
- CODE_MODE_ENABLED=true # Enable Code Mode
269
- CODE_MODE_CACHE_TTL=60000 # Tool list cache TTL (ms)
252
+ LOAD_SHEDDING_ENABLED=true
253
+ LOAD_SHEDDING_HEAP_THRESHOLD=0.85
270
254
  ```
271
255
 
272
- See [Token Optimization Guide](documentation/token-optimization.md#phase-0-mcp-code-mode-96-reduction-for-mcp-tools) and [Tools Documentation](documentation/tools.md#mcp-code-mode-token-optimization) for details.
256
+ ### Admin Hot-Reload (no restart needed)
257
+ ```bash
258
+ curl -X POST http://localhost:8081/v1/admin/reload
259
+ ```
273
260
 
274
261
  ---
275
262
 
276
- ## Deployment Options
263
+ ## Installation Methods
277
264
 
278
- **One-line install (recommended)**
265
+ **NPM (recommended)**
266
+ ```bash
267
+ npm install -g lynkr
268
+ ```
269
+
270
+ **One-line installer**
279
271
  ```bash
280
272
  curl -fsSL https://raw.githubusercontent.com/Fast-Editor/Lynkr/main/install.sh | bash
281
273
  ```
282
274
 
283
- **NPM**
275
+ **Homebrew**
284
276
  ```bash
285
- npm install -g lynkr && lynkr start
277
+ brew tap vishalveerareddy123/lynkr
278
+ brew install lynkr
286
279
  ```
287
280
 
288
281
  **Docker**
289
282
  ```bash
283
+ git clone https://github.com/Fast-Editor/Lynkr.git
284
+ cd Lynkr
290
285
  docker-compose up -d
291
286
  ```
292
287
 
293
- **Git Clone**
288
+ **From source**
294
289
  ```bash
295
290
  git clone https://github.com/Fast-Editor/Lynkr.git
296
- cd Lynkr && npm install && cp .env.example .env
291
+ cd Lynkr
292
+ npm install
293
+ cp .env.example .env
297
294
  npm start
298
295
  ```
299
296
 
300
- **Homebrew**
301
- ```bash
302
- brew tap vishalveerareddy123/lynkr
303
- brew install lynkr
304
- ```
305
-
306
297
  ---
307
298
 
308
299
  ## Documentation
@@ -310,54 +301,59 @@ brew install lynkr
310
301
  | Guide | Description |
311
302
  |-------|-------------|
312
303
  | [Installation](documentation/installation.md) | All installation methods |
313
- | [Provider Config](documentation/providers.md) | Setup for all 12+ providers |
314
- | [Claude Code CLI](documentation/claude-code-cli.md) | Detailed Claude Code integration |
315
- | [Codex CLI](documentation/codex-cli.md) | Codex config.toml setup |
316
- | [OpenClaw](documentation/openclaw-integration.md) | OpenClaw integration with tier routing |
317
- | [Cursor IDE](documentation/cursor-integration.md) | Cursor integration + troubleshooting |
318
- | [Embeddings](documentation/embeddings.md) | @Codebase semantic search (4 options) |
319
- | [Token Optimization](documentation/token-optimization.md) | 60-80% cost reduction strategies |
320
- | [Memory System](documentation/memory-system.md) | Titans-inspired long-term memory |
321
- | [Tools & Execution](documentation/tools.md) | Tool calling and execution modes |
322
- | [Smart Routing](documentation/routing.md) | Complexity-based model routing |
323
- | [Docker Deployment](documentation/docker.md) | docker-compose with GPU support |
324
- | [Production Hardening](documentation/production.md) | Circuit breakers, metrics, load shedding |
325
- | [API Reference](documentation/api.md) | All endpoints and formats |
304
+ | [Provider Setup](documentation/providers.md) | Configuration for all 12+ providers |
305
+ | [Claude Code](documentation/claude-code-cli.md) | Claude Code CLI integration |
306
+ | [Cursor IDE](documentation/cursor-integration.md) | Cursor setup + troubleshooting |
307
+ | [Codex CLI](documentation/codex-cli.md) | Codex configuration |
308
+ | [Tier Routing](documentation/routing.md) | Smart model routing by complexity |
309
+ | [Token Optimization](documentation/token-optimization.md) | 60-80% cost reduction |
326
310
  | [Troubleshooting](documentation/troubleshooting.md) | Common issues and solutions |
327
- | [FAQ](documentation/faq.md) | Frequently asked questions |
311
+ | [API Reference](documentation/api.md) | REST API endpoints |
312
+ | [Production](documentation/production.md) | Enterprise deployment |
328
313
 
329
314
  ---
330
315
 
331
- ## Troubleshooting
316
+ ## Cost Comparison
332
317
 
333
- | Issue | Solution |
334
- |-------|----------|
335
- | Same response for all queries | Disable semantic cache: `SEMANTIC_CACHE_ENABLED=false` |
336
- | Tool calls not executing | Increase threshold: `POLICY_TOOL_LOOP_THRESHOLD=15` |
337
- | Slow first request | Keep Ollama loaded: `OLLAMA_KEEP_ALIVE=24h` |
338
- | Connection refused | Ensure Lynkr is running: `lynkr start` |
318
+ | Scenario | Direct Anthropic | Lynkr + Ollama | Lynkr + OpenRouter |
319
+ |----------|-----------------|----------------|-------------------|
320
+ | Daily coding (8h) | $10-30/day | **$0 (free)** | $2-8/day |
321
+ | Monthly (heavy use) | $300-900 | **$0** | $60-240 |
322
+
323
+ With tier routing + token optimization: **additional 60-80% savings** on cloud providers.
339
324
 
340
325
  ---
341
326
 
342
- ## Contributing
327
+ ## Why Lynkr vs Alternatives
328
+
329
+ | Feature | Lynkr | LiteLLM | OpenRouter | PortKey |
330
+ |---------|-------|---------|-----------|---------|
331
+ | **Setup** | `npm install -g lynkr` | Python + Docker + Postgres | Account signup | Docker stack |
332
+ | **Claude Code native** | ✅ Drop-in | ⚠️ Requires config | ❌ | ⚠️ Partial |
333
+ | **Cursor native** | ✅ Drop-in | ⚠️ Partial | ❌ | ⚠️ Partial |
334
+ | **Local models** | Ollama, llama.cpp, LM Studio, MLX | Ollama only | ❌ | ❌ |
335
+ | **Tier routing** | Auto complexity-based | ❌ Manual | Cost-based only | ❌ Manual |
336
+ | **Token optimization** | 60-80% built-in | ❌ | ❌ | Cache only |
337
+ | **Self-hosted** | ✅ Node.js only | ✅ Python stack | ❌ SaaS | ✅ Docker |
338
+ | **Dependencies** | Node.js 20+ | Python, Prisma, PostgreSQL | None | Docker, Python |
343
339
 
344
- We welcome contributions. See the [Contributing Guide](documentation/contributing.md) and [Testing Guide](documentation/testing.md).
340
+ **Lynkr's edge:** Purpose-built for AI coding tools. Zero-config for Claude Code, Cursor, and Codex. Installs in one command, runs anywhere Node.js runs.
345
341
 
346
342
  ---
347
343
 
348
- ## License
344
+ ## Community
349
345
 
350
- Apache 2.0 — See [LICENSE](LICENSE).
346
+ - [GitHub Discussions](https://github.com/Fast-Editor/Lynkr/discussions) — Ask questions
347
+ - [Report Issues](https://github.com/Fast-Editor/Lynkr/issues) — Bug reports
348
+ - [NPM Package](https://www.npmjs.com/package/lynkr) — Official releases
349
+ - [DeepWiki](https://deepwiki.com/vishalveerareddy123/Lynkr) — AI-powered docs
351
350
 
352
351
  ---
353
352
 
354
- ## Community
353
+ ## License
355
354
 
356
- - [GitHub Discussions](https://github.com/Fast-Editor/Lynkr/discussions) — Questions and tips
357
- - [Report Issues](https://github.com/Fast-Editor/Lynkr/issues) — Bug reports and feature requests
358
- - [NPM Package](https://www.npmjs.com/package/lynkr) — Official package
359
- - [DeepWiki](https://deepwiki.com/vishalveerareddy123/Lynkr) — AI-powered docs search
355
+ Apache 2.0 — See [LICENSE](LICENSE).
360
356
 
361
357
  ---
362
358
 
363
- **Built by [Vishal Veera Reddy](https://github.com/vishalveerareddy123) — for developers who want control over their AI tools.**
359
+ **Built by [Vishal Veera Reddy](https://github.com/vishalveerareddy123) for developers who want control over their AI tools.**
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "lynkr",
3
- "version": "9.1.9",
3
+ "version": "9.2.0",
4
4
  "description": "Self-hosted Claude Code & Cursor proxy with Databricks,AWS BedRock,Azure adapters, openrouter, Ollama,llamacpp,LM Studio, workspace tooling, and MCP integration.",
5
5
  "main": "index.js",
6
6
  "bin": {